Fix multi rank finetune for ASR #7684

titu1994 · 2023-10-11T02:34:26Z

What does this PR do ?

Fix multi node fine-tuning script by adding rank based delay

Collection: [ASR]

Changelog

Add some delay to other ranks for the pretrained model call

Before your PR is "Ready for review"

Pre checks:

Make sure you read and followed Contributor guidelines
Did you write any new necessary tests?
Did you add or update any necessary documentation?
Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
- Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

New Feature
Bugfix
Documentation

nithinraok · 2023-10-11T17:45:29Z

examples/asr/speech_to_text_finetune.py

+            wait_time = int(cfg.get('exp_manager', {}).get('seconds_to_sleep', 60))
+            if wait_time < 60:
+                wait_time = 60
+                logging.info("Sleeping for at least 60 seconds to wait for model download to finish.")


lets keep this out of if case and show the seconds if passed through cfg?

Sounds good

Signed-off-by: smajumdar <titu1994@gmail.com>

* Fix multi rank finetune for ASR Signed-off-by: smajumdar <titu1994@gmail.com> * Actually add time Signed-off-by: smajumdar <titu1994@gmail.com> * Actually add time Signed-off-by: smajumdar <titu1994@gmail.com> --------- Signed-off-by: smajumdar <titu1994@gmail.com>

* Fix multi rank finetune for ASR * Actually add time * Actually add time --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com>

* Fix multi rank finetune for ASR * Actually add time * Actually add time --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Piotr Żelasko <petezor@gmail.com>

* Hotfix (#7501) (#7568) Signed-off-by: Jan Baczek <jbaczek@nvidia.com> Co-authored-by: jbaczek <45043825+jbaczek@users.noreply.github.com> * Avoid duplicated checkpoint save (#7555) (#7566) Signed-off-by: Mikołaj Błaż <mblaz@nvidia.com> Co-authored-by: mikolajblaz <mikolajblaz@users.noreply.github.com> * Cache FP8 weight and transpose only at the first micro-batch in each validation and test routine (#7470) (#7483) * Cache weight and transpose only in the first batch in all training, val, and test runs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Add an option to disable manual GC in validation (#7467) (#7476) Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> * Remove PUBLICATIONS.md, point to github.io NeMo page instead (#7694) (#7695) * update publications section to point to blog website page * add hyphen * use double backquotes for code formatting --------- Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Signed-off-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> * Fix multi rank finetune for ASR (#7684) (#7699) * Fix multi rank finetune for ASR * Actually add time * Actually add time --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> * Update docs: readme, getting started, ASR intro (#7679) * [TTS] Add dataset to path of logged artifacts (#7462) * [TTS] Add dataset to path of logged artifacts Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Revert axis name back to Audio Frames Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * move install info to INSTALLATION.md Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * tidy up links Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix sft dataset truncation (#7464) * Add fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Automatic Lip Reading Recognition (ALR) - ASR/CV (Visual ASR) (#7330) * striding_conv1d_k5 and dw_striding_conv1d_k5 subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * transpose conv1d inputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update subsampling.py change striding_conv1d_k5 to striding_conv1d Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * video manifest Signed-off-by: mburchi <maxime.burchi@gmail.com> * add collection classes Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test_step_outputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * clean references Signed-off-by: mburchi <maxime.burchi@gmail.com> * freeze unfreeze transcribe cv models Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest get_full_path bug Signed-off-by: mburchi <maxime.burchi@gmail.com> * update for PR Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * guard torchvision Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * _video_speech_collate_fn in cv/data/video_to_text.py Signed-off-by: mburchi <maxime.burchi@gmail.com> * add self.out = None to asr subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv -> multimodal/speech_cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: mburchi <maxime.burchi@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * HF StarCoder to NeMo conversion script (#7421) * Script to convert HF StarCoder checkpoint to NeMo Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * StarCoder conversion test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Fix test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Catch up with save_to changes Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Don't abbreviate args for clarity Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Configurable precision: BF16 vs FP32 Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix bug when loading dist ckpt in peft (#7452) Signed-off-by: Hongbin Liu <hongbinl@nvidia.com> Co-authored-by: Hongbin Liu <hongbinl@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix adding positional embeddings in-place in transformer module (#7440) Signed-off-by: Tamerlan Tabolov <tktabolov@gmail.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix (#7478) Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add sleep (#7498) (#7499) * add sleep * add sleep onto config instead * add comment --------- Signed-off-by: Gerald Shen <geshen@nvidia.com> Co-authored-by: Gerald Shen <119401249+gshennvm@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix exp manager check for sleep (#7503) (#7504) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.accelerator=auto from None. (#7492) (#7493) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [doc] fix broken link (#7481) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Read audio as int32 to avoid flac read errors (#7477) * [TTS] Read audio as int32 to avoid flac read errors Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add comment about read failures Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS (#7409) * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS * Train 'AISHELL-3' dataset with multi-speakers Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update get_data.py update copyright header Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * Update get_data.py added a disclaimer Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add new configuration file for AISHELL3 with multispeaker of fastpitch Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * dllogger - log on rank 0 only (#7513) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix TTS FastPitch tutorial (#7494) (#7516) * Fix --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix get_dist() tensor dimension (#7506) (#7515) Signed-off-by: Jocelyn Huang <jocelynh@nvidia.com> Co-authored-by: Jocelyn <jocelynh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: specify trainer.strategy=auto when devices=1 (#7509) (#7512) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix (#7511) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Fix FastPitch data prep tutorial (#7524) Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add italian tokenization (#7486) * add italian tokenization Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more ipa lexicon it Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix error deletion Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Replace None strategy with auto in tutorial notebooks (#7521) (#7527) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * unpin setuptools (#7534) (#7535) Signed-off-by: fayejf <36722593+fayejf@users.noreply.github.com> Co-authored-by: fayejf <36722593+fayejf@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * remove auto generated examples (#7510) * explicitly remove autogenerated examples for data parallel evaluation Signed-off-by: arendu <adithyare@nvidia.com> * mark autogenrated and remove it for test Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add the `strategy` argument to `MegatronGPTModel.generate()` (#7264) It is passed as an explicit argument rather than through `**strategy_args` so as to ensure someone cannot accidentally pass other arguments that would end up being ignored. It is a keyword-only argument to ensure that if in the future we want to update the signature to `**strategy_args`, we can do it without breaking code. Signed-off-by: Olivier Delalleau <507137+odelalleau@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix PTL2.0 related ASR bugs in r1.21.0: Val metrics logging, None dataloader issue (#7531) (#7533) * fix none dataloader issue ptl2 * ptl2.0 logging fixes for rnnt_models --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * gpus -> devices (#7542) (#7545) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Update FFMPEG version to fix issue with torchaudio (#7551) (#7553) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * PEFT GPT & T5 Refactor (#7308) * initial implementation of add_adapters API * correct type hint * Add config in add_adapters for save and load (@author bobchen) * Remove AdapterConfig to avoid import error * Add AdaterConfig back and move adaptermixin to sft model * Add NLPSaveRestoreConnector as default in NLPModel.restore_from * Add restore_from_nemo_with_adapter and test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * rename t5 file and classes to be consistent with GPT * add t5 sft dataset * add support for single-file format with T5SFTDataset * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Various small changes to make T5 SFT work like GPT SFT * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add adapter evaluation test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add MultiAdaterConfig for ia3 and fix builder issue * Make ptuning for T5SFTModel work using mixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add IA3_Adapter for AdapterName * Add adapter name for ptuning and attention adapter * Make test script GPT/T5 agnostic * Add layer selection feature * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Integrate adapter name and config * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update gpt peft tuning script to new API * add t5 peft tuning script with new API * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix IA3 layer selection issue * Override state_dict on SFT model instead of mixin * Add load adapter by adapter config * move peft config map away from example script * auto get config from nemo adapter * Move PEFTConfig to new file * fix ckpt save/load for t5 * name change: add_adapters -> add_adapter * variable name change * update t5 script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix t5 issues * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add weight tying * update gpt tuning script * PEFT-API proposal * Fix according to comments * update tuning scripts * move merge_cfg_with to mixin class since it applies to both gpt and t5 and requires the model class for restore * Add mcore_gpt support for NLPAdapterMixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix typo * variable name change to distinguish "peft" and "adapter" * override `load_adapters` to support `add_adapter` name change * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update tuning and eval script for adapter save/load * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add Ptuning on first stage only * add lora tutorial for review * Fix layer selection for mcore * add landing page * fix resume training Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add mcore condition in sharded_state_dict to make sft work * Update lora_tutorial.md First edit of this file for PEFT documentation for NeMO Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * rename Adapter to AttentionAdapter to avoid confusion in doc * Change load_adapters to load .nemo * add quick start guide * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add load_adapters with .ckpt * Remove setup_complete changes in load_adapters * update landing page * remove typo * Updated quick_start.md per Chen Cui Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * Add inference config merger and tutorial * Add doc string for NLPAdapterModelMixin and deprecated warning on MegatronGPTPEFTModel * add supported_methods.md and update other documentations * Update supported_methods.md minor updates. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Update landing_page.md minor update. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Modify doc string for NLPAdapterModelMixin * Add doc string add_adapters in NLPAdapterModelMixin * rename canonical adapters * remove mcore hard dependency * [PATCH] move microbatch calculator to nemo from apex * remove apex dependency in gpt and t5 sft models * remove apex dependency in gpt model * render doc strings * fix * Add missing virtual_tokens on ptuning * fix docstrings * update gpt-style model coverage in docs * update docstring * Remove pdb * add lightning_fabric to make docstring rendering work * Add Ptuning missing key * try docstring rendering * Fix ptuning issue * update gpt t5 peft tuning and eval scripts * typos * update eval config * fix bug relating to apex dependency removal * typo * make predict step behave the same as test step * make lora tutorial work in notebook * cosmetics * update yaml scripts * mcore_gpt attribute optional * typo * update eval scripts and fix T5 eval bugs * add NLPDDPStrategyNotebook and trainer builder logic to use it * update lora notebook to use new trainer builder * fix microbatch calculator bug for inference after training * Convert markdown files to RST and incorporate with doc * typo * revise language * remove extra cell * remove unnecessary inheritance * remove old tests * move layer selection default so logging messages make sense * remove `save_adapters` as adapter weights are saved automatically during training * initialize weights from a checkpoint instead of randomly * multiple fields can form a context (#7147) * list of context fields and flexible prompt template Signed-off-by: arendu <adithya.r@gmail.com> * list of fields for context Signed-off-by: arendu <adithya.r@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add multiple truncation fields and middle truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Compatible to old ckpt Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix tokenize detokenize issue Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove detokenization, add truncation augmentation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Resolve comments Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove unused import Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert eos Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add tokenizer space_sensitive attribute Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix error Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix erorr and use re Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Change assert logic Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Follow adi suggestion Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove merge function Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add example and comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove context_key and add comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove random truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix template none Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> * revert config changes * remove accidental breakpoint * support TP>1 loading * infer adapter type from checkpoint in during eval * breakup add adapter * enable interpolation of train_ds and validation_ds * update metric calc script to conform to single-file eval format * remove extraneous print * update lora notebook for updated merge_inference_cfg * Update nlp_adapter_mixins.py variable name change Signed-off-by: Chen Cui <chcui@nvidia.com> * turn off grad scaler for PP to match old scripts * remove PEFTSaveRestoreConnector since functionality all covered by the new mixin class * remove resume_from_checkpoint check since covered in #7335 * revert changes made in eval config interpolation * more interpolation * typo * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * remove dup line Signed-off-by: Chen Cui <chcui@nvidia.com> * code style warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix config mistake Signed-off-by: Chen Cui <chcui@nvidia.com> * add copyright header Signed-off-by: Chen Cui <chcui@nvidia.com> * fix code check warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert changes to remove apex dependency (mixed apex+nemo microbatch calculator broke some CI tests) Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * consolidate peft and sft scripts Signed-off-by: Chen Cui <chcui@nvidia.com> * update CI tests Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * notebook branch points to main to prepare for merge Signed-off-by: Chen Cui <chcui@nvidia.com> * fix gpt and t5 validation with any metric other than loss Signed-off-by: Chen Cui <chcui@nvidia.com> * support pre-extracted checkpoints Signed-off-by: Chen Cui <chcui@nvidia.com> --------- Signed-off-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Chen Cui <chcui@nvidia.com> Co-authored-by: Chen Cui <chcui@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Marc Romeyn <marcromeyn@gmail.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Yuanzhe Dong <yudong@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix a typo (#7496) Signed-off-by: BestJuly <chntaoli@163.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] remove curly braces from ${BRANCH} in jupyer notebook cell. (#7554) (#7560) * remove curly braces. * remove installation of pynini. --------- Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add youtube embed url (#7570) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Remap speakers to continuous range of speaker_id for dataset AISHELL3 (#7536) * Remap speakers to continuous range of speaker_id for dataset AISHELL3 * Add new key/value pair to record raw speaker for AISHELL3 dataset Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix validation_step_outputs initialization for multi-dataloader (#7546) (#7572) * added correct validation_step_outputs initialization for mutli-dataloader * changed kernel for display * Update logic for validation and test step outputs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert multidataloader changes in multilang ASR notebook --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append output of val step to self.validation_step_outputs (#7530) (#7532) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] fixed trainer's accelerator and strategy. (#7569) (#7574) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append val/test output to instance variable in EncDecSpeakerLabelModel (#7562) (#7573) * Append val/test output to the instance variable in EncDecSpeakerLabelModel * Handle test case in evaluation_step * Replace type with isinstance --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix CustomProgressBar for resume (#7427) (#7522) * Fix CustomProgress Bar for resume and multiple epochs * Edit num_training_batches * Use max_steps as total for progress bar for resume * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix typos in nfa and speech enhancement tutorials (#7580) (#7583) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add strategy as ddp_find_unused_parameters_true for glue_benchmark.py (#7454) (#7461) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * update strategy (#7577) (#7578) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix typos (#7581) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Change hifigan finetune strategy to ddp_find_unused_parameters_true (#7579) (#7584) * Change strategy to auto --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [BugFix] Add missing quotes for auto strategy in tutorial notebooks (#7541) (#7548) * Add missing quotes for auto strategy * Revert trainer.gpus to trainer.devices in Self_Supervised_Pre_Training.ipynb --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add build os key (#7596) (#7599) * add build os key * add tools * update to stable version --------- Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * StarCoder SFT test + bump PyT NGC image to 23.09 (#7540) * Add SFT StarCoder test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Remove _modify_config call as it is covered in load_from_nemo just below Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Test with pyt:23.09 container Signed-off-by: Jan Lasek <janek.lasek@gmail.com> --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * defaults changed (#7600) * defaults changed Signed-off-by: arendu <adithyare@nvidia.com> * typo Signed-off-by: arendu <adithyare@nvidia.com> * update Signed-off-by: arendu <adithyare@nvidia.com> --------- Signed-off-by: arendu <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add ItalianPhonemesTokenizer (#7587) * add ItalianPhonemesTokenizer Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix Italian phonemes Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * best ckpt fix (#7564) (#7588) Signed-off-by: dimapihtar <dpihtar@gmail.com> Co-authored-by: Dmytro Pykhtar <37850217+dimapihtar@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add files via upload (#7598) specifies the branch Signed-off-by: George <37293288+Jorjeous@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix validation in G2PModel and ThutmoseTaggerModel (#7597) (#7606) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Bound transformers version in requirements (#7620) Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix llama2 70b lora tuning bug (#7622) * fix llama2 70b lora tuning bug Signed-off-by: Chen Cui <chcui@nvidia.com> * Update peft_config.py brackets Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> --------- Signed-off-by: Chen Cui <chcui@nvidia.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix import error no module name model_utils (#7629) Signed-off-by: Mehadi Hasan Menon <mehadihasan80@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add fc large ls models (#7641) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao Koluguri <nithinraok> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.gpus, trainer.strategy, trainer.accelerator (#7621) (#7642) * [TTS] bugfix for Tacotron2 tutorial due to PTL 2.0 * trainer.gpus -> trainer.devices * fixed related tutorial bugs --------- Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix ssl models ptl monitor val through logging (#7608) (#7614) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Co-authored-by: Eric Harper <complex451@gmail.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix metrics for SE tutorial (#7604) (#7612) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Co-authored-by: anteju <108555623+anteju@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add ddp_find_unused_parameters=True and change accelerator to auto (#7623) (#7644) * Add ddp_find_unused_parameters=True and change acclerator to auto * Add ddp_find_unused_parameters True for normalization_as_tagging_train.py --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix py3.11 dataclasses issue (#7616) * Fix py3.11 dataclasses issue (#7582) * Update ASR configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Update TTS configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Guard MeCab and Ipadic Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix scripts Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Update name to ConfidenceMethodConfig Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Fix changes to confidence measure Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix issues with Dockerfile (#7650) (#7652) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] RNN-T greedy decoding max_frames fix for alignment and confidence (#7635) * decoding and test fix Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] Fix type error in jasper (#7636) (#7653) Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: Ryan Langman <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Add STFT and SI-SDR loss to audio codec recipe (#7468) * [TTS] Add STFT and SI-SDR loss to audio codec recipe Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix STFT resolution Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix training metric logging Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add docstring to mel and stft losses Signed-off-by: Ryan <rlangman@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add outline of asr quickstart info to asr/intro.rst Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add CLI, LM and real-time transcription sections Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Create per.py (#7538) * Move model precision copy (#7336) * move cfg precision set to megatron base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * remove copy from other models Signed-off-by: Maanu Grover <maanug@nvidia.com> * modify attribute not arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix gpt model test for ptl 2.0 Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename function and add docstring Signed-off-by: Maanu Grover <maanug@nvidia.com> * replace precision to dtype conditionals with func call Signed-off-by: Maanu Grover <maanug@nvidia.com> * unnecessary function and cfg reset Signed-off-by: Maanu Grover <maanug@nvidia.com> * set default value Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix precision lookup in a few more places Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename mapping function Signed-off-by: Maanu Grover <maanug@nvidia.com> * ununsed import Signed-off-by: Maanu Grover <maanug@nvidia.com> * save torch datatype to model Signed-off-by: Maanu Grover <maanug@nvidia.com> * set weights precision wrt amp o2 Signed-off-by: Maanu Grover <maanug@nvidia.com> * Revert "set weights precision wrt amp o2" This reverts commit 313a4bfe5eb69d771a6d2433898c0685836aef5c. Signed-off-by: Maanu Grover <maanug@nvidia.com> * revert half precision at inference attempt Signed-off-by: Maanu Grover <maanug@nvidia.com> * move autocast dtype to base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * move params dtype to base model, enable fp16 O2 inf Signed-off-by: Maanu Grover <maanug@nvidia.com> * unused imports Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix PEFT checkpoint loading (#7388) * Fix PEFT checkpoint loading Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use distributed optimizer support for multiple dtypes (#7359) * Update distopt wrapper with multiple dtype support Remove manual handling of separate FP32 optimizer. Signed-off-by: Tim Moon <tmoon@nvidia.com> * Use distopt support for contiguous buffers with multiple dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Fix typo Signed-off-by: Tim Moon <tmoon@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Separate distopt buckets for first GPT layer and non-overlapped params Signed-off-by: Tim Moon <tmoon@nvidia.com> * Add distopt logic for int dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit Signed-off-by: Tim Moon <tmoon@nvidia.com> * Remove unused variables Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit in README and Jenkensfile Signed-off-by: Tim Moon <tmoon@nvidia.com> * Debug Dockerfile and Jenkinsfile Signed-off-by: Tim Moon <tmoon@nvidia.com> --------- Signed-off-by: Tim Moon <tmoon@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * minor fix for llama ckpt conversion script (#7387) * minor fix for llama ckpt conversion script Signed-off-by: Jason Wang <jasonwan@nvidia.com> * Update Jenkinsfile Signed-off-by: Jason Wang <jasonwan@nvidia.com> * remove fast_swiglu configuration Signed-off-by: Jason Wang <jasonwan@nvidia.com> --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix wrong calling of librosa.get_duration() in notebook (#7376) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [PATCH] PEFT import mcore (#7393) * [PATCH] PEFT import mcore Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Create per.py Script for calculation Punctuation Error Rate and related rates (correct rate, deletions rate, etc.) Signed-off-by: Sasha Meister <117230141+ssh-meister@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Added a callback for logging initial data (#7384) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update Core Commit (#7402) * Update Core Commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * update commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use cfg attribute in bert (#7394) * use cfg attribute instead of arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * use torch_dtype in place of cfg.precision Signed-off-by: Maanu Grover <maanug@nvidia.com> * move precision copy before super constructor Signed-off-by: Maanu Grover <maanug@nvidia.com> * use trainer arg Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add support for bias conversion in Swiglu models (#7386) * Add support for bias conversion in Swiglu models Signed-off-by: smajumdar <titu1994@gmail.com> * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * Fix issue with missing tokenizer Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update save_to and restore_from for dist checkpointing (#7343) * add dist ckpt to save to, in progress Signed-off-by: eharper <eharper@nvidia.com> * move dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * clean up Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update restore from, need to figure out how to initialize distributed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * launch distrib if needed when restoring dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * when using mcore we can change tp pp on the fly Signed-off-by: eharper <eharper@nvidia.com> * add load_from_checkpoint support for dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update llama convert script to save dist .nemo Signed-off-by: eharper <eharper@nvidia.com> * fix load dist ckpt Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup TE TP groups if needed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup te tp groups if needed Signed-off-by: eharper <eharper@nvidia.com> * remove import Signed-off-by: eharper <eharper@nvidia.com> --------- Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * fix forward for with mcore=false (#7403) Signed-off-by: Jimmy Zhang <jiemingz@nvidia.com> Co-authored-by: Jimmy Zhang <jiemingz@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix logging to remove 's/it' from progress bar in Megatron models and add train_step_timing (#7374) * Add CustomProgressBar class to exp_manager and trainer callbacks Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix the progress bar to reflect total microbatch cnt Signed-off-by: Abhishree <abhishreetm@gmail.com> * Modify CustomProgressBar class 1) Modify CustomProgressBar class to update progress bar per global_step instead of per microbatch 2) Add the callback to other megatron training/finetuning files that are not using MegatronTrainerBuilder Signed-off-by: Abhishree <abhishreetm@gmail.com> * Add CustomProgressBar callback to tuning files Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Set Activation Checkpointing Defaults (#7404) * Set Activation Checkpointing Defaults Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * check for None Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * make loss mask default to false (#7407) Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add dummy userbuffer config files (#7408) Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * add missing ubconf files (#7412) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * New tutorial on Speech Data Explorer (#7405) * Added Google Colab based tutorial on Speech Data Explorer Signed-off-by: George Zelenfroynd <gzelenfroind@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update ptl training ckpt conversion script to work with dist ckpt (#7416) * update ptl convert script Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * don't break legacy Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: eharper <eharper@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Allow disabling sanity checking when num_sanity_val_steps=0 (#7413) * Allow disabling sanity checking when num_sanity_val_steps=0 Signed-off-by: Abhishree <abhishreetm@gmail.com> * Update num_sanity_val_steps to be a multiple of num_microbatches Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add comprehensive error messages (#7261) Signed-off-by: Anton Peganov <apeganov@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * check NEMO_PATH (#7418) Signed-off-by: Nikolay Karpov <karpnv@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * layer selection for ia3 (#7417) * layer selection for ia3 Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix missing pip package 'einops' (#7397) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of pyaudio in Google Colab (#7396) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update README.md: output_path --> output_manifest_filepath (#7442) Signed-off-by: Samuele Cornell <cornellsamuele@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add rope dynamic linear scaling (#7437) * Add dynamic linear scaling Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Yang Zhang <yzhang123@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix None dataloader issue in PTL2.0 (#7455) * Fix None dataloader issue in PTL2.0 Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [ASR] Confidence measure -> method renames (#7434) * measure -> method Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add steps for document of getting dataset 'SF Bilingual Speech' (#7378) * Add steps for document of getting dataset 'SF Bilingual Speech' Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update datasets.rst added a link from a tutorial demonstrating detailed data prep steps. Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * RNN-T confidence and alignment bugfix (#7381) * new frame_confidence and alignments lists are now always created after the while loop Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * tests added Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix resume from checkpoint in exp_manager (#7424) (#7426) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix checking of cuda/cpu device for inputs of Decoder (#7444) * Fix checking of cuda/cpu device for inputs of Decoder Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update tacotron2.py Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of ljspeech's get_data.py (#7430) * Fix failure of ljspeech's get_data.py Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Fix audio codec type checks (#7373) * [TTS] Fix audio codec type checks Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix audio codec tests Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * […

) * Hotfix (#7501) (#7568) Signed-off-by: Jan Baczek <jbaczek@nvidia.com> Co-authored-by: jbaczek <45043825+jbaczek@users.noreply.github.com> * Avoid duplicated checkpoint save (#7555) (#7566) Signed-off-by: Mikołaj Błaż <mblaz@nvidia.com> Co-authored-by: mikolajblaz <mikolajblaz@users.noreply.github.com> * Cache FP8 weight and transpose only at the first micro-batch in each validation and test routine (#7470) (#7483) * Cache weight and transpose only in the first batch in all training, val, and test runs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Add an option to disable manual GC in validation (#7467) (#7476) Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> * Remove PUBLICATIONS.md, point to github.io NeMo page instead (#7694) (#7695) * update publications section to point to blog website page * add hyphen * use double backquotes for code formatting --------- Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Signed-off-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> * Fix multi rank finetune for ASR (#7684) (#7699) * Fix multi rank finetune for ASR * Actually add time * Actually add time --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> * Update docs: readme, getting started, ASR intro (#7679) * [TTS] Add dataset to path of logged artifacts (#7462) * [TTS] Add dataset to path of logged artifacts Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Revert axis name back to Audio Frames Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * move install info to INSTALLATION.md Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * tidy up links Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix sft dataset truncation (#7464) * Add fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Automatic Lip Reading Recognition (ALR) - ASR/CV (Visual ASR) (#7330) * striding_conv1d_k5 and dw_striding_conv1d_k5 subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * transpose conv1d inputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update subsampling.py change striding_conv1d_k5 to striding_conv1d Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * video manifest Signed-off-by: mburchi <maxime.burchi@gmail.com> * add collection classes Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test_step_outputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * clean references Signed-off-by: mburchi <maxime.burchi@gmail.com> * freeze unfreeze transcribe cv models Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest get_full_path bug Signed-off-by: mburchi <maxime.burchi@gmail.com> * update for PR Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * guard torchvision Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * _video_speech_collate_fn in cv/data/video_to_text.py Signed-off-by: mburchi <maxime.burchi@gmail.com> * add self.out = None to asr subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv -> multimodal/speech_cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: mburchi <maxime.burchi@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * HF StarCoder to NeMo conversion script (#7421) * Script to convert HF StarCoder checkpoint to NeMo Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * StarCoder conversion test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Fix test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Catch up with save_to changes Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Don't abbreviate args for clarity Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Configurable precision: BF16 vs FP32 Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix bug when loading dist ckpt in peft (#7452) Signed-off-by: Hongbin Liu <hongbinl@nvidia.com> Co-authored-by: Hongbin Liu <hongbinl@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix adding positional embeddings in-place in transformer module (#7440) Signed-off-by: Tamerlan Tabolov <tktabolov@gmail.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix (#7478) Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add sleep (#7498) (#7499) * add sleep * add sleep onto config instead * add comment --------- Signed-off-by: Gerald Shen <geshen@nvidia.com> Co-authored-by: Gerald Shen <119401249+gshennvm@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix exp manager check for sleep (#7503) (#7504) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.accelerator=auto from None. (#7492) (#7493) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [doc] fix broken link (#7481) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Read audio as int32 to avoid flac read errors (#7477) * [TTS] Read audio as int32 to avoid flac read errors Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add comment about read failures Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS (#7409) * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS * Train 'AISHELL-3' dataset with multi-speakers Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update get_data.py update copyright header Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * Update get_data.py added a disclaimer Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add new configuration file for AISHELL3 with multispeaker of fastpitch Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * dllogger - log on rank 0 only (#7513) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix TTS FastPitch tutorial (#7494) (#7516) * Fix --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix get_dist() tensor dimension (#7506) (#7515) Signed-off-by: Jocelyn Huang <jocelynh@nvidia.com> Co-authored-by: Jocelyn <jocelynh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: specify trainer.strategy=auto when devices=1 (#7509) (#7512) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix (#7511) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Fix FastPitch data prep tutorial (#7524) Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add italian tokenization (#7486) * add italian tokenization Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more ipa lexicon it Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix error deletion Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Replace None strategy with auto in tutorial notebooks (#7521) (#7527) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * unpin setuptools (#7534) (#7535) Signed-off-by: fayejf <36722593+fayejf@users.noreply.github.com> Co-authored-by: fayejf <36722593+fayejf@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * remove auto generated examples (#7510) * explicitly remove autogenerated examples for data parallel evaluation Signed-off-by: arendu <adithyare@nvidia.com> * mark autogenrated and remove it for test Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add the `strategy` argument to `MegatronGPTModel.generate()` (#7264) It is passed as an explicit argument rather than through `**strategy_args` so as to ensure someone cannot accidentally pass other arguments that would end up being ignored. It is a keyword-only argument to ensure that if in the future we want to update the signature to `**strategy_args`, we can do it without breaking code. Signed-off-by: Olivier Delalleau <507137+odelalleau@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix PTL2.0 related ASR bugs in r1.21.0: Val metrics logging, None dataloader issue (#7531) (#7533) * fix none dataloader issue ptl2 * ptl2.0 logging fixes for rnnt_models --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * gpus -> devices (#7542) (#7545) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Update FFMPEG version to fix issue with torchaudio (#7551) (#7553) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * PEFT GPT & T5 Refactor (#7308) * initial implementation of add_adapters API * correct type hint * Add config in add_adapters for save and load (@author bobchen) * Remove AdapterConfig to avoid import error * Add AdaterConfig back and move adaptermixin to sft model * Add NLPSaveRestoreConnector as default in NLPModel.restore_from * Add restore_from_nemo_with_adapter and test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * rename t5 file and classes to be consistent with GPT * add t5 sft dataset * add support for single-file format with T5SFTDataset * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Various small changes to make T5 SFT work like GPT SFT * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add adapter evaluation test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add MultiAdaterConfig for ia3 and fix builder issue * Make ptuning for T5SFTModel work using mixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add IA3_Adapter for AdapterName * Add adapter name for ptuning and attention adapter * Make test script GPT/T5 agnostic * Add layer selection feature * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Integrate adapter name and config * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update gpt peft tuning script to new API * add t5 peft tuning script with new API * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix IA3 layer selection issue * Override state_dict on SFT model instead of mixin * Add load adapter by adapter config * move peft config map away from example script * auto get config from nemo adapter * Move PEFTConfig to new file * fix ckpt save/load for t5 * name change: add_adapters -> add_adapter * variable name change * update t5 script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix t5 issues * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add weight tying * update gpt tuning script * PEFT-API proposal * Fix according to comments * update tuning scripts * move merge_cfg_with to mixin class since it applies to both gpt and t5 and requires the model class for restore * Add mcore_gpt support for NLPAdapterMixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix typo * variable name change to distinguish "peft" and "adapter" * override `load_adapters` to support `add_adapter` name change * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update tuning and eval script for adapter save/load * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add Ptuning on first stage only * add lora tutorial for review * Fix layer selection for mcore * add landing page * fix resume training Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add mcore condition in sharded_state_dict to make sft work * Update lora_tutorial.md First edit of this file for PEFT documentation for NeMO Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * rename Adapter to AttentionAdapter to avoid confusion in doc * Change load_adapters to load .nemo * add quick start guide * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add load_adapters with .ckpt * Remove setup_complete changes in load_adapters * update landing page * remove typo * Updated quick_start.md per Chen Cui Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * Add inference config merger and tutorial * Add doc string for NLPAdapterModelMixin and deprecated warning on MegatronGPTPEFTModel * add supported_methods.md and update other documentations * Update supported_methods.md minor updates. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Update landing_page.md minor update. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Modify doc string for NLPAdapterModelMixin * Add doc string add_adapters in NLPAdapterModelMixin * rename canonical adapters * remove mcore hard dependency * [PATCH] move microbatch calculator to nemo from apex * remove apex dependency in gpt and t5 sft models * remove apex dependency in gpt model * render doc strings * fix * Add missing virtual_tokens on ptuning * fix docstrings * update gpt-style model coverage in docs * update docstring * Remove pdb * add lightning_fabric to make docstring rendering work * Add Ptuning missing key * try docstring rendering * Fix ptuning issue * update gpt t5 peft tuning and eval scripts * typos * update eval config * fix bug relating to apex dependency removal * typo * make predict step behave the same as test step * make lora tutorial work in notebook * cosmetics * update yaml scripts * mcore_gpt attribute optional * typo * update eval scripts and fix T5 eval bugs * add NLPDDPStrategyNotebook and trainer builder logic to use it * update lora notebook to use new trainer builder * fix microbatch calculator bug for inference after training * Convert markdown files to RST and incorporate with doc * typo * revise language * remove extra cell * remove unnecessary inheritance * remove old tests * move layer selection default so logging messages make sense * remove `save_adapters` as adapter weights are saved automatically during training * initialize weights from a checkpoint instead of randomly * multiple fields can form a context (#7147) * list of context fields and flexible prompt template Signed-off-by: arendu <adithya.r@gmail.com> * list of fields for context Signed-off-by: arendu <adithya.r@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add multiple truncation fields and middle truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Compatible to old ckpt Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix tokenize detokenize issue Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove detokenization, add truncation augmentation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Resolve comments Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove unused import Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert eos Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add tokenizer space_sensitive attribute Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix error Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix erorr and use re Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Change assert logic Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Follow adi suggestion Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove merge function Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add example and comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove context_key and add comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove random truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix template none Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> * revert config changes * remove accidental breakpoint * support TP>1 loading * infer adapter type from checkpoint in during eval * breakup add adapter * enable interpolation of train_ds and validation_ds * update metric calc script to conform to single-file eval format * remove extraneous print * update lora notebook for updated merge_inference_cfg * Update nlp_adapter_mixins.py variable name change Signed-off-by: Chen Cui <chcui@nvidia.com> * turn off grad scaler for PP to match old scripts * remove PEFTSaveRestoreConnector since functionality all covered by the new mixin class * remove resume_from_checkpoint check since covered in #7335 * revert changes made in eval config interpolation * more interpolation * typo * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * remove dup line Signed-off-by: Chen Cui <chcui@nvidia.com> * code style warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix config mistake Signed-off-by: Chen Cui <chcui@nvidia.com> * add copyright header Signed-off-by: Chen Cui <chcui@nvidia.com> * fix code check warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert changes to remove apex dependency (mixed apex+nemo microbatch calculator broke some CI tests) Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * consolidate peft and sft scripts Signed-off-by: Chen Cui <chcui@nvidia.com> * update CI tests Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * notebook branch points to main to prepare for merge Signed-off-by: Chen Cui <chcui@nvidia.com> * fix gpt and t5 validation with any metric other than loss Signed-off-by: Chen Cui <chcui@nvidia.com> * support pre-extracted checkpoints Signed-off-by: Chen Cui <chcui@nvidia.com> --------- Signed-off-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Chen Cui <chcui@nvidia.com> Co-authored-by: Chen Cui <chcui@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Marc Romeyn <marcromeyn@gmail.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Yuanzhe Dong <yudong@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix a typo (#7496) Signed-off-by: BestJuly <chntaoli@163.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] remove curly braces from ${BRANCH} in jupyer notebook cell. (#7554) (#7560) * remove curly braces. * remove installation of pynini. --------- Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add youtube embed url (#7570) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Remap speakers to continuous range of speaker_id for dataset AISHELL3 (#7536) * Remap speakers to continuous range of speaker_id for dataset AISHELL3 * Add new key/value pair to record raw speaker for AISHELL3 dataset Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix validation_step_outputs initialization for multi-dataloader (#7546) (#7572) * added correct validation_step_outputs initialization for mutli-dataloader * changed kernel for display * Update logic for validation and test step outputs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert multidataloader changes in multilang ASR notebook --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append output of val step to self.validation_step_outputs (#7530) (#7532) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] fixed trainer's accelerator and strategy. (#7569) (#7574) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append val/test output to instance variable in EncDecSpeakerLabelModel (#7562) (#7573) * Append val/test output to the instance variable in EncDecSpeakerLabelModel * Handle test case in evaluation_step * Replace type with isinstance --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix CustomProgressBar for resume (#7427) (#7522) * Fix CustomProgress Bar for resume and multiple epochs * Edit num_training_batches * Use max_steps as total for progress bar for resume * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix typos in nfa and speech enhancement tutorials (#7580) (#7583) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add strategy as ddp_find_unused_parameters_true for glue_benchmark.py (#7454) (#7461) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * update strategy (#7577) (#7578) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix typos (#7581) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Change hifigan finetune strategy to ddp_find_unused_parameters_true (#7579) (#7584) * Change strategy to auto --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [BugFix] Add missing quotes for auto strategy in tutorial notebooks (#7541) (#7548) * Add missing quotes for auto strategy * Revert trainer.gpus to trainer.devices in Self_Supervised_Pre_Training.ipynb --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add build os key (#7596) (#7599) * add build os key * add tools * update to stable version --------- Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * StarCoder SFT test + bump PyT NGC image to 23.09 (#7540) * Add SFT StarCoder test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Remove _modify_config call as it is covered in load_from_nemo just below Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Test with pyt:23.09 container Signed-off-by: Jan Lasek <janek.lasek@gmail.com> --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * defaults changed (#7600) * defaults changed Signed-off-by: arendu <adithyare@nvidia.com> * typo Signed-off-by: arendu <adithyare@nvidia.com> * update Signed-off-by: arendu <adithyare@nvidia.com> --------- Signed-off-by: arendu <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add ItalianPhonemesTokenizer (#7587) * add ItalianPhonemesTokenizer Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix Italian phonemes Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * best ckpt fix (#7564) (#7588) Signed-off-by: dimapihtar <dpihtar@gmail.com> Co-authored-by: Dmytro Pykhtar <37850217+dimapihtar@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add files via upload (#7598) specifies the branch Signed-off-by: George <37293288+Jorjeous@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix validation in G2PModel and ThutmoseTaggerModel (#7597) (#7606) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Bound transformers version in requirements (#7620) Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix llama2 70b lora tuning bug (#7622) * fix llama2 70b lora tuning bug Signed-off-by: Chen Cui <chcui@nvidia.com> * Update peft_config.py brackets Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> --------- Signed-off-by: Chen Cui <chcui@nvidia.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix import error no module name model_utils (#7629) Signed-off-by: Mehadi Hasan Menon <mehadihasan80@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add fc large ls models (#7641) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao Koluguri <nithinraok> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.gpus, trainer.strategy, trainer.accelerator (#7621) (#7642) * [TTS] bugfix for Tacotron2 tutorial due to PTL 2.0 * trainer.gpus -> trainer.devices * fixed related tutorial bugs --------- Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix ssl models ptl monitor val through logging (#7608) (#7614) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Co-authored-by: Eric Harper <complex451@gmail.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix metrics for SE tutorial (#7604) (#7612) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Co-authored-by: anteju <108555623+anteju@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add ddp_find_unused_parameters=True and change accelerator to auto (#7623) (#7644) * Add ddp_find_unused_parameters=True and change acclerator to auto * Add ddp_find_unused_parameters True for normalization_as_tagging_train.py --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix py3.11 dataclasses issue (#7616) * Fix py3.11 dataclasses issue (#7582) * Update ASR configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Update TTS configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Guard MeCab and Ipadic Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix scripts Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Update name to ConfidenceMethodConfig Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Fix changes to confidence measure Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix issues with Dockerfile (#7650) (#7652) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] RNN-T greedy decoding max_frames fix for alignment and confidence (#7635) * decoding and test fix Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] Fix type error in jasper (#7636) (#7653) Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: Ryan Langman <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Add STFT and SI-SDR loss to audio codec recipe (#7468) * [TTS] Add STFT and SI-SDR loss to audio codec recipe Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix STFT resolution Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix training metric logging Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add docstring to mel and stft losses Signed-off-by: Ryan <rlangman@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add outline of asr quickstart info to asr/intro.rst Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add CLI, LM and real-time transcription sections Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Create per.py (#7538) * Move model precision copy (#7336) * move cfg precision set to megatron base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * remove copy from other models Signed-off-by: Maanu Grover <maanug@nvidia.com> * modify attribute not arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix gpt model test for ptl 2.0 Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename function and add docstring Signed-off-by: Maanu Grover <maanug@nvidia.com> * replace precision to dtype conditionals with func call Signed-off-by: Maanu Grover <maanug@nvidia.com> * unnecessary function and cfg reset Signed-off-by: Maanu Grover <maanug@nvidia.com> * set default value Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix precision lookup in a few more places Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename mapping function Signed-off-by: Maanu Grover <maanug@nvidia.com> * ununsed import Signed-off-by: Maanu Grover <maanug@nvidia.com> * save torch datatype to model Signed-off-by: Maanu Grover <maanug@nvidia.com> * set weights precision wrt amp o2 Signed-off-by: Maanu Grover <maanug@nvidia.com> * Revert "set weights precision wrt amp o2" This reverts commit 313a4bfe5eb69d771a6d2433898c0685836aef5c. Signed-off-by: Maanu Grover <maanug@nvidia.com> * revert half precision at inference attempt Signed-off-by: Maanu Grover <maanug@nvidia.com> * move autocast dtype to base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * move params dtype to base model, enable fp16 O2 inf Signed-off-by: Maanu Grover <maanug@nvidia.com> * unused imports Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix PEFT checkpoint loading (#7388) * Fix PEFT checkpoint loading Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use distributed optimizer support for multiple dtypes (#7359) * Update distopt wrapper with multiple dtype support Remove manual handling of separate FP32 optimizer. Signed-off-by: Tim Moon <tmoon@nvidia.com> * Use distopt support for contiguous buffers with multiple dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Fix typo Signed-off-by: Tim Moon <tmoon@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Separate distopt buckets for first GPT layer and non-overlapped params Signed-off-by: Tim Moon <tmoon@nvidia.com> * Add distopt logic for int dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit Signed-off-by: Tim Moon <tmoon@nvidia.com> * Remove unused variables Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit in README and Jenkensfile Signed-off-by: Tim Moon <tmoon@nvidia.com> * Debug Dockerfile and Jenkinsfile Signed-off-by: Tim Moon <tmoon@nvidia.com> --------- Signed-off-by: Tim Moon <tmoon@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * minor fix for llama ckpt conversion script (#7387) * minor fix for llama ckpt conversion script Signed-off-by: Jason Wang <jasonwan@nvidia.com> * Update Jenkinsfile Signed-off-by: Jason Wang <jasonwan@nvidia.com> * remove fast_swiglu configuration Signed-off-by: Jason Wang <jasonwan@nvidia.com> --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix wrong calling of librosa.get_duration() in notebook (#7376) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [PATCH] PEFT import mcore (#7393) * [PATCH] PEFT import mcore Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Create per.py Script for calculation Punctuation Error Rate and related rates (correct rate, deletions rate, etc.) Signed-off-by: Sasha Meister <117230141+ssh-meister@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Added a callback for logging initial data (#7384) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update Core Commit (#7402) * Update Core Commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * update commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use cfg attribute in bert (#7394) * use cfg attribute instead of arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * use torch_dtype in place of cfg.precision Signed-off-by: Maanu Grover <maanug@nvidia.com> * move precision copy before super constructor Signed-off-by: Maanu Grover <maanug@nvidia.com> * use trainer arg Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add support for bias conversion in Swiglu models (#7386) * Add support for bias conversion in Swiglu models Signed-off-by: smajumdar <titu1994@gmail.com> * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * Fix issue with missing tokenizer Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update save_to and restore_from for dist checkpointing (#7343) * add dist ckpt to save to, in progress Signed-off-by: eharper <eharper@nvidia.com> * move dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * clean up Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update restore from, need to figure out how to initialize distributed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * launch distrib if needed when restoring dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * when using mcore we can change tp pp on the fly Signed-off-by: eharper <eharper@nvidia.com> * add load_from_checkpoint support for dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update llama convert script to save dist .nemo Signed-off-by: eharper <eharper@nvidia.com> * fix load dist ckpt Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup TE TP groups if needed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup te tp groups if needed Signed-off-by: eharper <eharper@nvidia.com> * remove import Signed-off-by: eharper <eharper@nvidia.com> --------- Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * fix forward for with mcore=false (#7403) Signed-off-by: Jimmy Zhang <jiemingz@nvidia.com> Co-authored-by: Jimmy Zhang <jiemingz@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix logging to remove 's/it' from progress bar in Megatron models and add train_step_timing (#7374) * Add CustomProgressBar class to exp_manager and trainer callbacks Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix the progress bar to reflect total microbatch cnt Signed-off-by: Abhishree <abhishreetm@gmail.com> * Modify CustomProgressBar class 1) Modify CustomProgressBar class to update progress bar per global_step instead of per microbatch 2) Add the callback to other megatron training/finetuning files that are not using MegatronTrainerBuilder Signed-off-by: Abhishree <abhishreetm@gmail.com> * Add CustomProgressBar callback to tuning files Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Set Activation Checkpointing Defaults (#7404) * Set Activation Checkpointing Defaults Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * check for None Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * make loss mask default to false (#7407) Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add dummy userbuffer config files (#7408) Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * add missing ubconf files (#7412) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * New tutorial on Speech Data Explorer (#7405) * Added Google Colab based tutorial on Speech Data Explorer Signed-off-by: George Zelenfroynd <gzelenfroind@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update ptl training ckpt conversion script to work with dist ckpt (#7416) * update ptl convert script Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * don't break legacy Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: eharper <eharper@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Allow disabling sanity checking when num_sanity_val_steps=0 (#7413) * Allow disabling sanity checking when num_sanity_val_steps=0 Signed-off-by: Abhishree <abhishreetm@gmail.com> * Update num_sanity_val_steps to be a multiple of num_microbatches Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add comprehensive error messages (#7261) Signed-off-by: Anton Peganov <apeganov@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * check NEMO_PATH (#7418) Signed-off-by: Nikolay Karpov <karpnv@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * layer selection for ia3 (#7417) * layer selection for ia3 Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix missing pip package 'einops' (#7397) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of pyaudio in Google Colab (#7396) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update README.md: output_path --> output_manifest_filepath (#7442) Signed-off-by: Samuele Cornell <cornellsamuele@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add rope dynamic linear scaling (#7437) * Add dynamic linear scaling Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Yang Zhang <yzhang123@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix None dataloader issue in PTL2.0 (#7455) * Fix None dataloader issue in PTL2.0 Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [ASR] Confidence measure -> method renames (#7434) * measure -> method Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add steps for document of getting dataset 'SF Bilingual Speech' (#7378) * Add steps for document of getting dataset 'SF Bilingual Speech' Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update datasets.rst added a link from a tutorial demonstrating detailed data prep steps. Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * RNN-T confidence and alignment bugfix (#7381) * new frame_confidence and alignments lists are now always created after the while loop Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * tests added Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix resume from checkpoint in exp_manager (#7424) (#7426) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix checking of cuda/cpu device for inputs of Decoder (#7444) * Fix checking of cuda/cpu device for inputs of Decoder Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update tacotron2.py Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of ljspeech's get_data.py (#7430) * Fix failure of ljspeech's get_data.py Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Fix audio codec type checks (#7373) * [TTS] Fix audio codec type checks Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix audio codec tests Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Add dataset to path of logged artifacts (#7462) * [TTS] Add dataset to path of logged artifacts Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Revert axis name back to Audio Frames Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix sft dataset truncation (#7464) * Add fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Automatic Lip Reading Recognition (ALR) - ASR/CV (Visual ASR) (#7330) * striding_conv1d_k5 and dw_striding_conv1d_k5 subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * transpose conv1d inputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update subsampling.py change striding_conv1d_k5 to striding_conv1d Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * video manifest Signed-off-by: mburchi <maxime.burchi@gmail.com> * add collection classes Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-… * fix onnx (#7703) (#7704) Signed-off-by: fayejf <fayejf07@gmail.com> Co-authored-by: fayejf <36722593+fayejf@users.noreply.github.com> * move core install to /workspace (#7706) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * Fix typo in audio codec config, encoder target (#7697) Signed-off-by: Ante Jukić <ajukic@nvidia.com> * Replace strategy='dp'/None with 'auto' (#7681) (#7696) * Add strategy=auto for None and dp * Change strategy from None to auto --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> * [ASR] Multichannel mask estimator with flex number of channels (#7317) * Adding a mask estimator which can …

* Hotfix (#7501) (#7568) Signed-off-by: Jan Baczek <jbaczek@nvidia.com> Co-authored-by: jbaczek <45043825+jbaczek@users.noreply.github.com> * Avoid duplicated checkpoint save (#7555) (#7566) Signed-off-by: Mikołaj Błaż <mblaz@nvidia.com> Co-authored-by: mikolajblaz <mikolajblaz@users.noreply.github.com> * Cache FP8 weight and transpose only at the first micro-batch in each validation and test routine (#7470) (#7483) * Cache weight and transpose only in the first batch in all training, val, and test runs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Add an option to disable manual GC in validation (#7467) (#7476) Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> * Remove PUBLICATIONS.md, point to github.io NeMo page instead (#7694) (#7695) * update publications section to point to blog website page * add hyphen * use double backquotes for code formatting --------- Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Signed-off-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> * Fix multi rank finetune for ASR (#7684) (#7699) * Fix multi rank finetune for ASR * Actually add time * Actually add time --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> * Update docs: readme, getting started, ASR intro (#7679) * [TTS] Add dataset to path of logged artifacts (#7462) * [TTS] Add dataset to path of logged artifacts Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Revert axis name back to Audio Frames Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * move install info to INSTALLATION.md Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * tidy up links Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix sft dataset truncation (#7464) * Add fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Automatic Lip Reading Recognition (ALR) - ASR/CV (Visual ASR) (#7330) * striding_conv1d_k5 and dw_striding_conv1d_k5 subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * transpose conv1d inputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update subsampling.py change striding_conv1d_k5 to striding_conv1d Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * video manifest Signed-off-by: mburchi <maxime.burchi@gmail.com> * add collection classes Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test_step_outputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * clean references Signed-off-by: mburchi <maxime.burchi@gmail.com> * freeze unfreeze transcribe cv models Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest get_full_path bug Signed-off-by: mburchi <maxime.burchi@gmail.com> * update for PR Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * guard torchvision Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * _video_speech_collate_fn in cv/data/video_to_text.py Signed-off-by: mburchi <maxime.burchi@gmail.com> * add self.out = None to asr subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv -> multimodal/speech_cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: mburchi <maxime.burchi@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * HF StarCoder to NeMo conversion script (#7421) * Script to convert HF StarCoder checkpoint to NeMo Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * StarCoder conversion test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Fix test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Catch up with save_to changes Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Don't abbreviate args for clarity Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Configurable precision: BF16 vs FP32 Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix bug when loading dist ckpt in peft (#7452) Signed-off-by: Hongbin Liu <hongbinl@nvidia.com> Co-authored-by: Hongbin Liu <hongbinl@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix adding positional embeddings in-place in transformer module (#7440) Signed-off-by: Tamerlan Tabolov <tktabolov@gmail.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix (#7478) Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add sleep (#7498) (#7499) * add sleep * add sleep onto config instead * add comment --------- Signed-off-by: Gerald Shen <geshen@nvidia.com> Co-authored-by: Gerald Shen <119401249+gshennvm@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix exp manager check for sleep (#7503) (#7504) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.accelerator=auto from None. (#7492) (#7493) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [doc] fix broken link (#7481) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Read audio as int32 to avoid flac read errors (#7477) * [TTS] Read audio as int32 to avoid flac read errors Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add comment about read failures Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS (#7409) * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS * Train 'AISHELL-3' dataset with multi-speakers Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update get_data.py update copyright header Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * Update get_data.py added a disclaimer Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add new configuration file for AISHELL3 with multispeaker of fastpitch Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * dllogger - log on rank 0 only (#7513) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix TTS FastPitch tutorial (#7494) (#7516) * Fix --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix get_dist() tensor dimension (#7506) (#7515) Signed-off-by: Jocelyn Huang <jocelynh@nvidia.com> Co-authored-by: Jocelyn <jocelynh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: specify trainer.strategy=auto when devices=1 (#7509) (#7512) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix (#7511) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Fix FastPitch data prep tutorial (#7524) Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add italian tokenization (#7486) * add italian tokenization Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more ipa lexicon it Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix error deletion Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Replace None strategy with auto in tutorial notebooks (#7521) (#7527) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * unpin setuptools (#7534) (#7535) Signed-off-by: fayejf <36722593+fayejf@users.noreply.github.com> Co-authored-by: fayejf <36722593+fayejf@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * remove auto generated examples (#7510) * explicitly remove autogenerated examples for data parallel evaluation Signed-off-by: arendu <adithyare@nvidia.com> * mark autogenrated and remove it for test Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add the `strategy` argument to `MegatronGPTModel.generate()` (#7264) It is passed as an explicit argument rather than through `**strategy_args` so as to ensure someone cannot accidentally pass other arguments that would end up being ignored. It is a keyword-only argument to ensure that if in the future we want to update the signature to `**strategy_args`, we can do it without breaking code. Signed-off-by: Olivier Delalleau <507137+odelalleau@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix PTL2.0 related ASR bugs in r1.21.0: Val metrics logging, None dataloader issue (#7531) (#7533) * fix none dataloader issue ptl2 * ptl2.0 logging fixes for rnnt_models --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * gpus -> devices (#7542) (#7545) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Update FFMPEG version to fix issue with torchaudio (#7551) (#7553) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * PEFT GPT & T5 Refactor (#7308) * initial implementation of add_adapters API * correct type hint * Add config in add_adapters for save and load (@author bobchen) * Remove AdapterConfig to avoid import error * Add AdaterConfig back and move adaptermixin to sft model * Add NLPSaveRestoreConnector as default in NLPModel.restore_from * Add restore_from_nemo_with_adapter and test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * rename t5 file and classes to be consistent with GPT * add t5 sft dataset * add support for single-file format with T5SFTDataset * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Various small changes to make T5 SFT work like GPT SFT * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add adapter evaluation test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add MultiAdaterConfig for ia3 and fix builder issue * Make ptuning for T5SFTModel work using mixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add IA3_Adapter for AdapterName * Add adapter name for ptuning and attention adapter * Make test script GPT/T5 agnostic * Add layer selection feature * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Integrate adapter name and config * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update gpt peft tuning script to new API * add t5 peft tuning script with new API * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix IA3 layer selection issue * Override state_dict on SFT model instead of mixin * Add load adapter by adapter config * move peft config map away from example script * auto get config from nemo adapter * Move PEFTConfig to new file * fix ckpt save/load for t5 * name change: add_adapters -> add_adapter * variable name change * update t5 script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix t5 issues * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add weight tying * update gpt tuning script * PEFT-API proposal * Fix according to comments * update tuning scripts * move merge_cfg_with to mixin class since it applies to both gpt and t5 and requires the model class for restore * Add mcore_gpt support for NLPAdapterMixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix typo * variable name change to distinguish "peft" and "adapter" * override `load_adapters` to support `add_adapter` name change * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update tuning and eval script for adapter save/load * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add Ptuning on first stage only * add lora tutorial for review * Fix layer selection for mcore * add landing page * fix resume training Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add mcore condition in sharded_state_dict to make sft work * Update lora_tutorial.md First edit of this file for PEFT documentation for NeMO Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * rename Adapter to AttentionAdapter to avoid confusion in doc * Change load_adapters to load .nemo * add quick start guide * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add load_adapters with .ckpt * Remove setup_complete changes in load_adapters * update landing page * remove typo * Updated quick_start.md per Chen Cui Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * Add inference config merger and tutorial * Add doc string for NLPAdapterModelMixin and deprecated warning on MegatronGPTPEFTModel * add supported_methods.md and update other documentations * Update supported_methods.md minor updates. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Update landing_page.md minor update. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Modify doc string for NLPAdapterModelMixin * Add doc string add_adapters in NLPAdapterModelMixin * rename canonical adapters * remove mcore hard dependency * [PATCH] move microbatch calculator to nemo from apex * remove apex dependency in gpt and t5 sft models * remove apex dependency in gpt model * render doc strings * fix * Add missing virtual_tokens on ptuning * fix docstrings * update gpt-style model coverage in docs * update docstring * Remove pdb * add lightning_fabric to make docstring rendering work * Add Ptuning missing key * try docstring rendering * Fix ptuning issue * update gpt t5 peft tuning and eval scripts * typos * update eval config * fix bug relating to apex dependency removal * typo * make predict step behave the same as test step * make lora tutorial work in notebook * cosmetics * update yaml scripts * mcore_gpt attribute optional * typo * update eval scripts and fix T5 eval bugs * add NLPDDPStrategyNotebook and trainer builder logic to use it * update lora notebook to use new trainer builder * fix microbatch calculator bug for inference after training * Convert markdown files to RST and incorporate with doc * typo * revise language * remove extra cell * remove unnecessary inheritance * remove old tests * move layer selection default so logging messages make sense * remove `save_adapters` as adapter weights are saved automatically during training * initialize weights from a checkpoint instead of randomly * multiple fields can form a context (#7147) * list of context fields and flexible prompt template Signed-off-by: arendu <adithya.r@gmail.com> * list of fields for context Signed-off-by: arendu <adithya.r@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add multiple truncation fields and middle truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Compatible to old ckpt Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix tokenize detokenize issue Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove detokenization, add truncation augmentation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Resolve comments Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove unused import Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert eos Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add tokenizer space_sensitive attribute Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix error Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix erorr and use re Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Change assert logic Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Follow adi suggestion Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove merge function Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add example and comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove context_key and add comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove random truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix template none Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> * revert config changes * remove accidental breakpoint * support TP>1 loading * infer adapter type from checkpoint in during eval * breakup add adapter * enable interpolation of train_ds and validation_ds * update metric calc script to conform to single-file eval format * remove extraneous print * update lora notebook for updated merge_inference_cfg * Update nlp_adapter_mixins.py variable name change Signed-off-by: Chen Cui <chcui@nvidia.com> * turn off grad scaler for PP to match old scripts * remove PEFTSaveRestoreConnector since functionality all covered by the new mixin class * remove resume_from_checkpoint check since covered in #7335 * revert changes made in eval config interpolation * more interpolation * typo * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * remove dup line Signed-off-by: Chen Cui <chcui@nvidia.com> * code style warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix config mistake Signed-off-by: Chen Cui <chcui@nvidia.com> * add copyright header Signed-off-by: Chen Cui <chcui@nvidia.com> * fix code check warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert changes to remove apex dependency (mixed apex+nemo microbatch calculator broke some CI tests) Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * consolidate peft and sft scripts Signed-off-by: Chen Cui <chcui@nvidia.com> * update CI tests Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * notebook branch points to main to prepare for merge Signed-off-by: Chen Cui <chcui@nvidia.com> * fix gpt and t5 validation with any metric other than loss Signed-off-by: Chen Cui <chcui@nvidia.com> * support pre-extracted checkpoints Signed-off-by: Chen Cui <chcui@nvidia.com> --------- Signed-off-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Chen Cui <chcui@nvidia.com> Co-authored-by: Chen Cui <chcui@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Marc Romeyn <marcromeyn@gmail.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Yuanzhe Dong <yudong@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix a typo (#7496) Signed-off-by: BestJuly <chntaoli@163.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] remove curly braces from ${BRANCH} in jupyer notebook cell. (#7554) (#7560) * remove curly braces. * remove installation of pynini. --------- Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add youtube embed url (#7570) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Remap speakers to continuous range of speaker_id for dataset AISHELL3 (#7536) * Remap speakers to continuous range of speaker_id for dataset AISHELL3 * Add new key/value pair to record raw speaker for AISHELL3 dataset Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix validation_step_outputs initialization for multi-dataloader (#7546) (#7572) * added correct validation_step_outputs initialization for mutli-dataloader * changed kernel for display * Update logic for validation and test step outputs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert multidataloader changes in multilang ASR notebook --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append output of val step to self.validation_step_outputs (#7530) (#7532) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] fixed trainer's accelerator and strategy. (#7569) (#7574) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append val/test output to instance variable in EncDecSpeakerLabelModel (#7562) (#7573) * Append val/test output to the instance variable in EncDecSpeakerLabelModel * Handle test case in evaluation_step * Replace type with isinstance --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix CustomProgressBar for resume (#7427) (#7522) * Fix CustomProgress Bar for resume and multiple epochs * Edit num_training_batches * Use max_steps as total for progress bar for resume * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix typos in nfa and speech enhancement tutorials (#7580) (#7583) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add strategy as ddp_find_unused_parameters_true for glue_benchmark.py (#7454) (#7461) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * update strategy (#7577) (#7578) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix typos (#7581) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Change hifigan finetune strategy to ddp_find_unused_parameters_true (#7579) (#7584) * Change strategy to auto --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [BugFix] Add missing quotes for auto strategy in tutorial notebooks (#7541) (#7548) * Add missing quotes for auto strategy * Revert trainer.gpus to trainer.devices in Self_Supervised_Pre_Training.ipynb --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add build os key (#7596) (#7599) * add build os key * add tools * update to stable version --------- Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * StarCoder SFT test + bump PyT NGC image to 23.09 (#7540) * Add SFT StarCoder test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Remove _modify_config call as it is covered in load_from_nemo just below Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Test with pyt:23.09 container Signed-off-by: Jan Lasek <janek.lasek@gmail.com> --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * defaults changed (#7600) * defaults changed Signed-off-by: arendu <adithyare@nvidia.com> * typo Signed-off-by: arendu <adithyare@nvidia.com> * update Signed-off-by: arendu <adithyare@nvidia.com> --------- Signed-off-by: arendu <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add ItalianPhonemesTokenizer (#7587) * add ItalianPhonemesTokenizer Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix Italian phonemes Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * best ckpt fix (#7564) (#7588) Signed-off-by: dimapihtar <dpihtar@gmail.com> Co-authored-by: Dmytro Pykhtar <37850217+dimapihtar@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add files via upload (#7598) specifies the branch Signed-off-by: George <37293288+Jorjeous@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix validation in G2PModel and ThutmoseTaggerModel (#7597) (#7606) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Bound transformers version in requirements (#7620) Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix llama2 70b lora tuning bug (#7622) * fix llama2 70b lora tuning bug Signed-off-by: Chen Cui <chcui@nvidia.com> * Update peft_config.py brackets Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> --------- Signed-off-by: Chen Cui <chcui@nvidia.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix import error no module name model_utils (#7629) Signed-off-by: Mehadi Hasan Menon <mehadihasan80@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add fc large ls models (#7641) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao Koluguri <nithinraok> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.gpus, trainer.strategy, trainer.accelerator (#7621) (#7642) * [TTS] bugfix for Tacotron2 tutorial due to PTL 2.0 * trainer.gpus -> trainer.devices * fixed related tutorial bugs --------- Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix ssl models ptl monitor val through logging (#7608) (#7614) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Co-authored-by: Eric Harper <complex451@gmail.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix metrics for SE tutorial (#7604) (#7612) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Co-authored-by: anteju <108555623+anteju@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add ddp_find_unused_parameters=True and change accelerator to auto (#7623) (#7644) * Add ddp_find_unused_parameters=True and change acclerator to auto * Add ddp_find_unused_parameters True for normalization_as_tagging_train.py --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix py3.11 dataclasses issue (#7616) * Fix py3.11 dataclasses issue (#7582) * Update ASR configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Update TTS configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Guard MeCab and Ipadic Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix scripts Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Update name to ConfidenceMethodConfig Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Fix changes to confidence measure Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix issues with Dockerfile (#7650) (#7652) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] RNN-T greedy decoding max_frames fix for alignment and confidence (#7635) * decoding and test fix Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] Fix type error in jasper (#7636) (#7653) Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: Ryan Langman <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Add STFT and SI-SDR loss to audio codec recipe (#7468) * [TTS] Add STFT and SI-SDR loss to audio codec recipe Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix STFT resolution Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix training metric logging Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add docstring to mel and stft losses Signed-off-by: Ryan <rlangman@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add outline of asr quickstart info to asr/intro.rst Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add CLI, LM and real-time transcription sections Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Create per.py (#7538) * Move model precision copy (#7336) * move cfg precision set to megatron base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * remove copy from other models Signed-off-by: Maanu Grover <maanug@nvidia.com> * modify attribute not arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix gpt model test for ptl 2.0 Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename function and add docstring Signed-off-by: Maanu Grover <maanug@nvidia.com> * replace precision to dtype conditionals with func call Signed-off-by: Maanu Grover <maanug@nvidia.com> * unnecessary function and cfg reset Signed-off-by: Maanu Grover <maanug@nvidia.com> * set default value Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix precision lookup in a few more places Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename mapping function Signed-off-by: Maanu Grover <maanug@nvidia.com> * ununsed import Signed-off-by: Maanu Grover <maanug@nvidia.com> * save torch datatype to model Signed-off-by: Maanu Grover <maanug@nvidia.com> * set weights precision wrt amp o2 Signed-off-by: Maanu Grover <maanug@nvidia.com> * Revert "set weights precision wrt amp o2" This reverts commit 313a4bfe5eb69d771a6d2433898c0685836aef5c. Signed-off-by: Maanu Grover <maanug@nvidia.com> * revert half precision at inference attempt Signed-off-by: Maanu Grover <maanug@nvidia.com> * move autocast dtype to base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * move params dtype to base model, enable fp16 O2 inf Signed-off-by: Maanu Grover <maanug@nvidia.com> * unused imports Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix PEFT checkpoint loading (#7388) * Fix PEFT checkpoint loading Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use distributed optimizer support for multiple dtypes (#7359) * Update distopt wrapper with multiple dtype support Remove manual handling of separate FP32 optimizer. Signed-off-by: Tim Moon <tmoon@nvidia.com> * Use distopt support for contiguous buffers with multiple dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Fix typo Signed-off-by: Tim Moon <tmoon@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Separate distopt buckets for first GPT layer and non-overlapped params Signed-off-by: Tim Moon <tmoon@nvidia.com> * Add distopt logic for int dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit Signed-off-by: Tim Moon <tmoon@nvidia.com> * Remove unused variables Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit in README and Jenkensfile Signed-off-by: Tim Moon <tmoon@nvidia.com> * Debug Dockerfile and Jenkinsfile Signed-off-by: Tim Moon <tmoon@nvidia.com> --------- Signed-off-by: Tim Moon <tmoon@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * minor fix for llama ckpt conversion script (#7387) * minor fix for llama ckpt conversion script Signed-off-by: Jason Wang <jasonwan@nvidia.com> * Update Jenkinsfile Signed-off-by: Jason Wang <jasonwan@nvidia.com> * remove fast_swiglu configuration Signed-off-by: Jason Wang <jasonwan@nvidia.com> --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix wrong calling of librosa.get_duration() in notebook (#7376) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [PATCH] PEFT import mcore (#7393) * [PATCH] PEFT import mcore Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Create per.py Script for calculation Punctuation Error Rate and related rates (correct rate, deletions rate, etc.) Signed-off-by: Sasha Meister <117230141+ssh-meister@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Added a callback for logging initial data (#7384) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update Core Commit (#7402) * Update Core Commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * update commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use cfg attribute in bert (#7394) * use cfg attribute instead of arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * use torch_dtype in place of cfg.precision Signed-off-by: Maanu Grover <maanug@nvidia.com> * move precision copy before super constructor Signed-off-by: Maanu Grover <maanug@nvidia.com> * use trainer arg Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add support for bias conversion in Swiglu models (#7386) * Add support for bias conversion in Swiglu models Signed-off-by: smajumdar <titu1994@gmail.com> * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * Fix issue with missing tokenizer Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update save_to and restore_from for dist checkpointing (#7343) * add dist ckpt to save to, in progress Signed-off-by: eharper <eharper@nvidia.com> * move dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * clean up Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update restore from, need to figure out how to initialize distributed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * launch distrib if needed when restoring dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * when using mcore we can change tp pp on the fly Signed-off-by: eharper <eharper@nvidia.com> * add load_from_checkpoint support for dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update llama convert script to save dist .nemo Signed-off-by: eharper <eharper@nvidia.com> * fix load dist ckpt Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup TE TP groups if needed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup te tp groups if needed Signed-off-by: eharper <eharper@nvidia.com> * remove import Signed-off-by: eharper <eharper@nvidia.com> --------- Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * fix forward for with mcore=false (#7403) Signed-off-by: Jimmy Zhang <jiemingz@nvidia.com> Co-authored-by: Jimmy Zhang <jiemingz@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix logging to remove 's/it' from progress bar in Megatron models and add train_step_timing (#7374) * Add CustomProgressBar class to exp_manager and trainer callbacks Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix the progress bar to reflect total microbatch cnt Signed-off-by: Abhishree <abhishreetm@gmail.com> * Modify CustomProgressBar class 1) Modify CustomProgressBar class to update progress bar per global_step instead of per microbatch 2) Add the callback to other megatron training/finetuning files that are not using MegatronTrainerBuilder Signed-off-by: Abhishree <abhishreetm@gmail.com> * Add CustomProgressBar callback to tuning files Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Set Activation Checkpointing Defaults (#7404) * Set Activation Checkpointing Defaults Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * check for None Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * make loss mask default to false (#7407) Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add dummy userbuffer config files (#7408) Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * add missing ubconf files (#7412) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * New tutorial on Speech Data Explorer (#7405) * Added Google Colab based tutorial on Speech Data Explorer Signed-off-by: George Zelenfroynd <gzelenfroind@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update ptl training ckpt conversion script to work with dist ckpt (#7416) * update ptl convert script Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * don't break legacy Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: eharper <eharper@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Allow disabling sanity checking when num_sanity_val_steps=0 (#7413) * Allow disabling sanity checking when num_sanity_val_steps=0 Signed-off-by: Abhishree <abhishreetm@gmail.com> * Update num_sanity_val_steps to be a multiple of num_microbatches Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add comprehensive error messages (#7261) Signed-off-by: Anton Peganov <apeganov@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * check NEMO_PATH (#7418) Signed-off-by: Nikolay Karpov <karpnv@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * layer selection for ia3 (#7417) * layer selection for ia3 Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix missing pip package 'einops' (#7397) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of pyaudio in Google Colab (#7396) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update README.md: output_path --> output_manifest_filepath (#7442) Signed-off-by: Samuele Cornell <cornellsamuele@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add rope dynamic linear scaling (#7437) * Add dynamic linear scaling Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Yang Zhang <yzhang123@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix None dataloader issue in PTL2.0 (#7455) * Fix None dataloader issue in PTL2.0 Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [ASR] Confidence measure -> method renames (#7434) * measure -> method Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add steps for document of getting dataset 'SF Bilingual Speech' (#7378) * Add steps for document of getting dataset 'SF Bilingual Speech' Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update datasets.rst added a link from a tutorial demonstrating detailed data prep steps. Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * RNN-T confidence and alignment bugfix (#7381) * new frame_confidence and alignments lists are now always created after the while loop Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * tests added Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix resume from checkpoint in exp_manager (#7424) (#7426) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix checking of cuda/cpu device for inputs of Decoder (#7444) * Fix checking of cuda/cpu device for inputs of Decoder Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update tacotron2.py Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of ljspeech's get_data.py (#7430) * Fix failure of ljspeech's get_data.py Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Fix audio codec type checks (#7373) * [TTS] Fix audio codec type checks Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix audio codec tests Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * […

) * Hotfix (#7501) (#7568) Signed-off-by: Jan Baczek <jbaczek@nvidia.com> Co-authored-by: jbaczek <45043825+jbaczek@users.noreply.github.com> * Avoid duplicated checkpoint save (#7555) (#7566) Signed-off-by: Mikołaj Błaż <mblaz@nvidia.com> Co-authored-by: mikolajblaz <mikolajblaz@users.noreply.github.com> * Cache FP8 weight and transpose only at the first micro-batch in each validation and test routine (#7470) (#7483) * Cache weight and transpose only in the first batch in all training, val, and test runs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Add an option to disable manual GC in validation (#7467) (#7476) Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> * Remove PUBLICATIONS.md, point to github.io NeMo page instead (#7694) (#7695) * update publications section to point to blog website page * add hyphen * use double backquotes for code formatting --------- Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Signed-off-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> * Fix multi rank finetune for ASR (#7684) (#7699) * Fix multi rank finetune for ASR * Actually add time * Actually add time --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> * Update docs: readme, getting started, ASR intro (#7679) * [TTS] Add dataset to path of logged artifacts (#7462) * [TTS] Add dataset to path of logged artifacts Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Revert axis name back to Audio Frames Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * move install info to INSTALLATION.md Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * tidy up links Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix sft dataset truncation (#7464) * Add fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Automatic Lip Reading Recognition (ALR) - ASR/CV (Visual ASR) (#7330) * striding_conv1d_k5 and dw_striding_conv1d_k5 subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * transpose conv1d inputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update subsampling.py change striding_conv1d_k5 to striding_conv1d Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * video manifest Signed-off-by: mburchi <maxime.burchi@gmail.com> * add collection classes Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test_step_outputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * clean references Signed-off-by: mburchi <maxime.burchi@gmail.com> * freeze unfreeze transcribe cv models Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest get_full_path bug Signed-off-by: mburchi <maxime.burchi@gmail.com> * update for PR Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * guard torchvision Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * _video_speech_collate_fn in cv/data/video_to_text.py Signed-off-by: mburchi <maxime.burchi@gmail.com> * add self.out = None to asr subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv -> multimodal/speech_cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: mburchi <maxime.burchi@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * HF StarCoder to NeMo conversion script (#7421) * Script to convert HF StarCoder checkpoint to NeMo Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * StarCoder conversion test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Fix test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Catch up with save_to changes Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Don't abbreviate args for clarity Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Configurable precision: BF16 vs FP32 Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix bug when loading dist ckpt in peft (#7452) Signed-off-by: Hongbin Liu <hongbinl@nvidia.com> Co-authored-by: Hongbin Liu <hongbinl@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix adding positional embeddings in-place in transformer module (#7440) Signed-off-by: Tamerlan Tabolov <tktabolov@gmail.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix (#7478) Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add sleep (#7498) (#7499) * add sleep * add sleep onto config instead * add comment --------- Signed-off-by: Gerald Shen <geshen@nvidia.com> Co-authored-by: Gerald Shen <119401249+gshennvm@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix exp manager check for sleep (#7503) (#7504) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.accelerator=auto from None. (#7492) (#7493) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [doc] fix broken link (#7481) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Read audio as int32 to avoid flac read errors (#7477) * [TTS] Read audio as int32 to avoid flac read errors Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add comment about read failures Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS (#7409) * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS * Train 'AISHELL-3' dataset with multi-speakers Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update get_data.py update copyright header Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * Update get_data.py added a disclaimer Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add new configuration file for AISHELL3 with multispeaker of fastpitch Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * dllogger - log on rank 0 only (#7513) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix TTS FastPitch tutorial (#7494) (#7516) * Fix --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix get_dist() tensor dimension (#7506) (#7515) Signed-off-by: Jocelyn Huang <jocelynh@nvidia.com> Co-authored-by: Jocelyn <jocelynh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: specify trainer.strategy=auto when devices=1 (#7509) (#7512) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix (#7511) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Fix FastPitch data prep tutorial (#7524) Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add italian tokenization (#7486) * add italian tokenization Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more ipa lexicon it Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix error deletion Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Replace None strategy with auto in tutorial notebooks (#7521) (#7527) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * unpin setuptools (#7534) (#7535) Signed-off-by: fayejf <36722593+fayejf@users.noreply.github.com> Co-authored-by: fayejf <36722593+fayejf@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * remove auto generated examples (#7510) * explicitly remove autogenerated examples for data parallel evaluation Signed-off-by: arendu <adithyare@nvidia.com> * mark autogenrated and remove it for test Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add the `strategy` argument to `MegatronGPTModel.generate()` (#7264) It is passed as an explicit argument rather than through `**strategy_args` so as to ensure someone cannot accidentally pass other arguments that would end up being ignored. It is a keyword-only argument to ensure that if in the future we want to update the signature to `**strategy_args`, we can do it without breaking code. Signed-off-by: Olivier Delalleau <507137+odelalleau@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix PTL2.0 related ASR bugs in r1.21.0: Val metrics logging, None dataloader issue (#7531) (#7533) * fix none dataloader issue ptl2 * ptl2.0 logging fixes for rnnt_models --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * gpus -> devices (#7542) (#7545) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Update FFMPEG version to fix issue with torchaudio (#7551) (#7553) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * PEFT GPT & T5 Refactor (#7308) * initial implementation of add_adapters API * correct type hint * Add config in add_adapters for save and load (@author bobchen) * Remove AdapterConfig to avoid import error * Add AdaterConfig back and move adaptermixin to sft model * Add NLPSaveRestoreConnector as default in NLPModel.restore_from * Add restore_from_nemo_with_adapter and test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * rename t5 file and classes to be consistent with GPT * add t5 sft dataset * add support for single-file format with T5SFTDataset * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Various small changes to make T5 SFT work like GPT SFT * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add adapter evaluation test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add MultiAdaterConfig for ia3 and fix builder issue * Make ptuning for T5SFTModel work using mixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add IA3_Adapter for AdapterName * Add adapter name for ptuning and attention adapter * Make test script GPT/T5 agnostic * Add layer selection feature * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Integrate adapter name and config * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update gpt peft tuning script to new API * add t5 peft tuning script with new API * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix IA3 layer selection issue * Override state_dict on SFT model instead of mixin * Add load adapter by adapter config * move peft config map away from example script * auto get config from nemo adapter * Move PEFTConfig to new file * fix ckpt save/load for t5 * name change: add_adapters -> add_adapter * variable name change * update t5 script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix t5 issues * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add weight tying * update gpt tuning script * PEFT-API proposal * Fix according to comments * update tuning scripts * move merge_cfg_with to mixin class since it applies to both gpt and t5 and requires the model class for restore * Add mcore_gpt support for NLPAdapterMixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix typo * variable name change to distinguish "peft" and "adapter" * override `load_adapters` to support `add_adapter` name change * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update tuning and eval script for adapter save/load * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add Ptuning on first stage only * add lora tutorial for review * Fix layer selection for mcore * add landing page * fix resume training Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add mcore condition in sharded_state_dict to make sft work * Update lora_tutorial.md First edit of this file for PEFT documentation for NeMO Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * rename Adapter to AttentionAdapter to avoid confusion in doc * Change load_adapters to load .nemo * add quick start guide * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add load_adapters with .ckpt * Remove setup_complete changes in load_adapters * update landing page * remove typo * Updated quick_start.md per Chen Cui Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * Add inference config merger and tutorial * Add doc string for NLPAdapterModelMixin and deprecated warning on MegatronGPTPEFTModel * add supported_methods.md and update other documentations * Update supported_methods.md minor updates. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Update landing_page.md minor update. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Modify doc string for NLPAdapterModelMixin * Add doc string add_adapters in NLPAdapterModelMixin * rename canonical adapters * remove mcore hard dependency * [PATCH] move microbatch calculator to nemo from apex * remove apex dependency in gpt and t5 sft models * remove apex dependency in gpt model * render doc strings * fix * Add missing virtual_tokens on ptuning * fix docstrings * update gpt-style model coverage in docs * update docstring * Remove pdb * add lightning_fabric to make docstring rendering work * Add Ptuning missing key * try docstring rendering * Fix ptuning issue * update gpt t5 peft tuning and eval scripts * typos * update eval config * fix bug relating to apex dependency removal * typo * make predict step behave the same as test step * make lora tutorial work in notebook * cosmetics * update yaml scripts * mcore_gpt attribute optional * typo * update eval scripts and fix T5 eval bugs * add NLPDDPStrategyNotebook and trainer builder logic to use it * update lora notebook to use new trainer builder * fix microbatch calculator bug for inference after training * Convert markdown files to RST and incorporate with doc * typo * revise language * remove extra cell * remove unnecessary inheritance * remove old tests * move layer selection default so logging messages make sense * remove `save_adapters` as adapter weights are saved automatically during training * initialize weights from a checkpoint instead of randomly * multiple fields can form a context (#7147) * list of context fields and flexible prompt template Signed-off-by: arendu <adithya.r@gmail.com> * list of fields for context Signed-off-by: arendu <adithya.r@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add multiple truncation fields and middle truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Compatible to old ckpt Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix tokenize detokenize issue Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove detokenization, add truncation augmentation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Resolve comments Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove unused import Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert eos Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add tokenizer space_sensitive attribute Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix error Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix erorr and use re Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Change assert logic Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Follow adi suggestion Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove merge function Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add example and comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove context_key and add comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove random truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix template none Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> * revert config changes * remove accidental breakpoint * support TP>1 loading * infer adapter type from checkpoint in during eval * breakup add adapter * enable interpolation of train_ds and validation_ds * update metric calc script to conform to single-file eval format * remove extraneous print * update lora notebook for updated merge_inference_cfg * Update nlp_adapter_mixins.py variable name change Signed-off-by: Chen Cui <chcui@nvidia.com> * turn off grad scaler for PP to match old scripts * remove PEFTSaveRestoreConnector since functionality all covered by the new mixin class * remove resume_from_checkpoint check since covered in #7335 * revert changes made in eval config interpolation * more interpolation * typo * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * remove dup line Signed-off-by: Chen Cui <chcui@nvidia.com> * code style warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix config mistake Signed-off-by: Chen Cui <chcui@nvidia.com> * add copyright header Signed-off-by: Chen Cui <chcui@nvidia.com> * fix code check warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert changes to remove apex dependency (mixed apex+nemo microbatch calculator broke some CI tests) Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * consolidate peft and sft scripts Signed-off-by: Chen Cui <chcui@nvidia.com> * update CI tests Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * notebook branch points to main to prepare for merge Signed-off-by: Chen Cui <chcui@nvidia.com> * fix gpt and t5 validation with any metric other than loss Signed-off-by: Chen Cui <chcui@nvidia.com> * support pre-extracted checkpoints Signed-off-by: Chen Cui <chcui@nvidia.com> --------- Signed-off-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Chen Cui <chcui@nvidia.com> Co-authored-by: Chen Cui <chcui@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Marc Romeyn <marcromeyn@gmail.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Yuanzhe Dong <yudong@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix a typo (#7496) Signed-off-by: BestJuly <chntaoli@163.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] remove curly braces from ${BRANCH} in jupyer notebook cell. (#7554) (#7560) * remove curly braces. * remove installation of pynini. --------- Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add youtube embed url (#7570) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Remap speakers to continuous range of speaker_id for dataset AISHELL3 (#7536) * Remap speakers to continuous range of speaker_id for dataset AISHELL3 * Add new key/value pair to record raw speaker for AISHELL3 dataset Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix validation_step_outputs initialization for multi-dataloader (#7546) (#7572) * added correct validation_step_outputs initialization for mutli-dataloader * changed kernel for display * Update logic for validation and test step outputs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert multidataloader changes in multilang ASR notebook --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append output of val step to self.validation_step_outputs (#7530) (#7532) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] fixed trainer's accelerator and strategy. (#7569) (#7574) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append val/test output to instance variable in EncDecSpeakerLabelModel (#7562) (#7573) * Append val/test output to the instance variable in EncDecSpeakerLabelModel * Handle test case in evaluation_step * Replace type with isinstance --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix CustomProgressBar for resume (#7427) (#7522) * Fix CustomProgress Bar for resume and multiple epochs * Edit num_training_batches * Use max_steps as total for progress bar for resume * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix typos in nfa and speech enhancement tutorials (#7580) (#7583) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add strategy as ddp_find_unused_parameters_true for glue_benchmark.py (#7454) (#7461) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * update strategy (#7577) (#7578) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix typos (#7581) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Change hifigan finetune strategy to ddp_find_unused_parameters_true (#7579) (#7584) * Change strategy to auto --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [BugFix] Add missing quotes for auto strategy in tutorial notebooks (#7541) (#7548) * Add missing quotes for auto strategy * Revert trainer.gpus to trainer.devices in Self_Supervised_Pre_Training.ipynb --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add build os key (#7596) (#7599) * add build os key * add tools * update to stable version --------- Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * StarCoder SFT test + bump PyT NGC image to 23.09 (#7540) * Add SFT StarCoder test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Remove _modify_config call as it is covered in load_from_nemo just below Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Test with pyt:23.09 container Signed-off-by: Jan Lasek <janek.lasek@gmail.com> --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * defaults changed (#7600) * defaults changed Signed-off-by: arendu <adithyare@nvidia.com> * typo Signed-off-by: arendu <adithyare@nvidia.com> * update Signed-off-by: arendu <adithyare@nvidia.com> --------- Signed-off-by: arendu <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add ItalianPhonemesTokenizer (#7587) * add ItalianPhonemesTokenizer Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix Italian phonemes Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * best ckpt fix (#7564) (#7588) Signed-off-by: dimapihtar <dpihtar@gmail.com> Co-authored-by: Dmytro Pykhtar <37850217+dimapihtar@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add files via upload (#7598) specifies the branch Signed-off-by: George <37293288+Jorjeous@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix validation in G2PModel and ThutmoseTaggerModel (#7597) (#7606) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Bound transformers version in requirements (#7620) Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix llama2 70b lora tuning bug (#7622) * fix llama2 70b lora tuning bug Signed-off-by: Chen Cui <chcui@nvidia.com> * Update peft_config.py brackets Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> --------- Signed-off-by: Chen Cui <chcui@nvidia.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix import error no module name model_utils (#7629) Signed-off-by: Mehadi Hasan Menon <mehadihasan80@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add fc large ls models (#7641) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao Koluguri <nithinraok> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.gpus, trainer.strategy, trainer.accelerator (#7621) (#7642) * [TTS] bugfix for Tacotron2 tutorial due to PTL 2.0 * trainer.gpus -> trainer.devices * fixed related tutorial bugs --------- Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix ssl models ptl monitor val through logging (#7608) (#7614) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Co-authored-by: Eric Harper <complex451@gmail.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix metrics for SE tutorial (#7604) (#7612) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Co-authored-by: anteju <108555623+anteju@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add ddp_find_unused_parameters=True and change accelerator to auto (#7623) (#7644) * Add ddp_find_unused_parameters=True and change acclerator to auto * Add ddp_find_unused_parameters True for normalization_as_tagging_train.py --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix py3.11 dataclasses issue (#7616) * Fix py3.11 dataclasses issue (#7582) * Update ASR configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Update TTS configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Guard MeCab and Ipadic Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix scripts Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Update name to ConfidenceMethodConfig Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Fix changes to confidence measure Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix issues with Dockerfile (#7650) (#7652) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] RNN-T greedy decoding max_frames fix for alignment and confidence (#7635) * decoding and test fix Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] Fix type error in jasper (#7636) (#7653) Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: Ryan Langman <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Add STFT and SI-SDR loss to audio codec recipe (#7468) * [TTS] Add STFT and SI-SDR loss to audio codec recipe Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix STFT resolution Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix training metric logging Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add docstring to mel and stft losses Signed-off-by: Ryan <rlangman@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add outline of asr quickstart info to asr/intro.rst Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add CLI, LM and real-time transcription sections Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Create per.py (#7538) * Move model precision copy (#7336) * move cfg precision set to megatron base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * remove copy from other models Signed-off-by: Maanu Grover <maanug@nvidia.com> * modify attribute not arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix gpt model test for ptl 2.0 Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename function and add docstring Signed-off-by: Maanu Grover <maanug@nvidia.com> * replace precision to dtype conditionals with func call Signed-off-by: Maanu Grover <maanug@nvidia.com> * unnecessary function and cfg reset Signed-off-by: Maanu Grover <maanug@nvidia.com> * set default value Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix precision lookup in a few more places Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename mapping function Signed-off-by: Maanu Grover <maanug@nvidia.com> * ununsed import Signed-off-by: Maanu Grover <maanug@nvidia.com> * save torch datatype to model Signed-off-by: Maanu Grover <maanug@nvidia.com> * set weights precision wrt amp o2 Signed-off-by: Maanu Grover <maanug@nvidia.com> * Revert "set weights precision wrt amp o2" This reverts commit 313a4bfe5eb69d771a6d2433898c0685836aef5c. Signed-off-by: Maanu Grover <maanug@nvidia.com> * revert half precision at inference attempt Signed-off-by: Maanu Grover <maanug@nvidia.com> * move autocast dtype to base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * move params dtype to base model, enable fp16 O2 inf Signed-off-by: Maanu Grover <maanug@nvidia.com> * unused imports Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix PEFT checkpoint loading (#7388) * Fix PEFT checkpoint loading Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use distributed optimizer support for multiple dtypes (#7359) * Update distopt wrapper with multiple dtype support Remove manual handling of separate FP32 optimizer. Signed-off-by: Tim Moon <tmoon@nvidia.com> * Use distopt support for contiguous buffers with multiple dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Fix typo Signed-off-by: Tim Moon <tmoon@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Separate distopt buckets for first GPT layer and non-overlapped params Signed-off-by: Tim Moon <tmoon@nvidia.com> * Add distopt logic for int dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit Signed-off-by: Tim Moon <tmoon@nvidia.com> * Remove unused variables Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit in README and Jenkensfile Signed-off-by: Tim Moon <tmoon@nvidia.com> * Debug Dockerfile and Jenkinsfile Signed-off-by: Tim Moon <tmoon@nvidia.com> --------- Signed-off-by: Tim Moon <tmoon@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * minor fix for llama ckpt conversion script (#7387) * minor fix for llama ckpt conversion script Signed-off-by: Jason Wang <jasonwan@nvidia.com> * Update Jenkinsfile Signed-off-by: Jason Wang <jasonwan@nvidia.com> * remove fast_swiglu configuration Signed-off-by: Jason Wang <jasonwan@nvidia.com> --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix wrong calling of librosa.get_duration() in notebook (#7376) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [PATCH] PEFT import mcore (#7393) * [PATCH] PEFT import mcore Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Create per.py Script for calculation Punctuation Error Rate and related rates (correct rate, deletions rate, etc.) Signed-off-by: Sasha Meister <117230141+ssh-meister@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Added a callback for logging initial data (#7384) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update Core Commit (#7402) * Update Core Commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * update commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use cfg attribute in bert (#7394) * use cfg attribute instead of arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * use torch_dtype in place of cfg.precision Signed-off-by: Maanu Grover <maanug@nvidia.com> * move precision copy before super constructor Signed-off-by: Maanu Grover <maanug@nvidia.com> * use trainer arg Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add support for bias conversion in Swiglu models (#7386) * Add support for bias conversion in Swiglu models Signed-off-by: smajumdar <titu1994@gmail.com> * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * Fix issue with missing tokenizer Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update save_to and restore_from for dist checkpointing (#7343) * add dist ckpt to save to, in progress Signed-off-by: eharper <eharper@nvidia.com> * move dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * clean up Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update restore from, need to figure out how to initialize distributed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * launch distrib if needed when restoring dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * when using mcore we can change tp pp on the fly Signed-off-by: eharper <eharper@nvidia.com> * add load_from_checkpoint support for dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update llama convert script to save dist .nemo Signed-off-by: eharper <eharper@nvidia.com> * fix load dist ckpt Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup TE TP groups if needed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup te tp groups if needed Signed-off-by: eharper <eharper@nvidia.com> * remove import Signed-off-by: eharper <eharper@nvidia.com> --------- Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * fix forward for with mcore=false (#7403) Signed-off-by: Jimmy Zhang <jiemingz@nvidia.com> Co-authored-by: Jimmy Zhang <jiemingz@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix logging to remove 's/it' from progress bar in Megatron models and add train_step_timing (#7374) * Add CustomProgressBar class to exp_manager and trainer callbacks Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix the progress bar to reflect total microbatch cnt Signed-off-by: Abhishree <abhishreetm@gmail.com> * Modify CustomProgressBar class 1) Modify CustomProgressBar class to update progress bar per global_step instead of per microbatch 2) Add the callback to other megatron training/finetuning files that are not using MegatronTrainerBuilder Signed-off-by: Abhishree <abhishreetm@gmail.com> * Add CustomProgressBar callback to tuning files Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Set Activation Checkpointing Defaults (#7404) * Set Activation Checkpointing Defaults Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * check for None Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * make loss mask default to false (#7407) Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add dummy userbuffer config files (#7408) Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * add missing ubconf files (#7412) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * New tutorial on Speech Data Explorer (#7405) * Added Google Colab based tutorial on Speech Data Explorer Signed-off-by: George Zelenfroynd <gzelenfroind@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update ptl training ckpt conversion script to work with dist ckpt (#7416) * update ptl convert script Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * don't break legacy Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: eharper <eharper@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Allow disabling sanity checking when num_sanity_val_steps=0 (#7413) * Allow disabling sanity checking when num_sanity_val_steps=0 Signed-off-by: Abhishree <abhishreetm@gmail.com> * Update num_sanity_val_steps to be a multiple of num_microbatches Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add comprehensive error messages (#7261) Signed-off-by: Anton Peganov <apeganov@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * check NEMO_PATH (#7418) Signed-off-by: Nikolay Karpov <karpnv@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * layer selection for ia3 (#7417) * layer selection for ia3 Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix missing pip package 'einops' (#7397) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of pyaudio in Google Colab (#7396) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update README.md: output_path --> output_manifest_filepath (#7442) Signed-off-by: Samuele Cornell <cornellsamuele@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add rope dynamic linear scaling (#7437) * Add dynamic linear scaling Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Yang Zhang <yzhang123@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix None dataloader issue in PTL2.0 (#7455) * Fix None dataloader issue in PTL2.0 Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [ASR] Confidence measure -> method renames (#7434) * measure -> method Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add steps for document of getting dataset 'SF Bilingual Speech' (#7378) * Add steps for document of getting dataset 'SF Bilingual Speech' Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update datasets.rst added a link from a tutorial demonstrating detailed data prep steps. Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * RNN-T confidence and alignment bugfix (#7381) * new frame_confidence and alignments lists are now always created after the while loop Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * tests added Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix resume from checkpoint in exp_manager (#7424) (#7426) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix checking of cuda/cpu device for inputs of Decoder (#7444) * Fix checking of cuda/cpu device for inputs of Decoder Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update tacotron2.py Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of ljspeech's get_data.py (#7430) * Fix failure of ljspeech's get_data.py Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Fix audio codec type checks (#7373) * [TTS] Fix audio codec type checks Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix audio codec tests Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * […

* Hotfix (#7501) (#7568) Signed-off-by: Jan Baczek <jbaczek@nvidia.com> Co-authored-by: jbaczek <45043825+jbaczek@users.noreply.github.com> * Avoid duplicated checkpoint save (#7555) (#7566) Signed-off-by: Mikołaj Błaż <mblaz@nvidia.com> Co-authored-by: mikolajblaz <mikolajblaz@users.noreply.github.com> * Cache FP8 weight and transpose only at the first micro-batch in each validation and test routine (#7470) (#7483) * Cache weight and transpose only in the first batch in all training, val, and test runs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Add an option to disable manual GC in validation (#7467) (#7476) Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> * Remove PUBLICATIONS.md, point to github.io NeMo page instead (#7694) (#7695) * update publications section to point to blog website page * add hyphen * use double backquotes for code formatting --------- Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Signed-off-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> * Fix multi rank finetune for ASR (#7684) (#7699) * Fix multi rank finetune for ASR * Actually add time * Actually add time --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> * Update docs: readme, getting started, ASR intro (#7679) * [TTS] Add dataset to path of logged artifacts (#7462) * [TTS] Add dataset to path of logged artifacts Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Revert axis name back to Audio Frames Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * move install info to INSTALLATION.md Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * tidy up links Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix sft dataset truncation (#7464) * Add fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Automatic Lip Reading Recognition (ALR) - ASR/CV (Visual ASR) (#7330) * striding_conv1d_k5 and dw_striding_conv1d_k5 subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * transpose conv1d inputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update subsampling.py change striding_conv1d_k5 to striding_conv1d Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * video manifest Signed-off-by: mburchi <maxime.burchi@gmail.com> * add collection classes Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test_step_outputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * clean references Signed-off-by: mburchi <maxime.burchi@gmail.com> * freeze unfreeze transcribe cv models Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest get_full_path bug Signed-off-by: mburchi <maxime.burchi@gmail.com> * update for PR Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * guard torchvision Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * _video_speech_collate_fn in cv/data/video_to_text.py Signed-off-by: mburchi <maxime.burchi@gmail.com> * add self.out = None to asr subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv -> multimodal/speech_cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: mburchi <maxime.burchi@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * HF StarCoder to NeMo conversion script (#7421) * Script to convert HF StarCoder checkpoint to NeMo Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * StarCoder conversion test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Fix test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Catch up with save_to changes Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Don't abbreviate args for clarity Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Configurable precision: BF16 vs FP32 Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix bug when loading dist ckpt in peft (#7452) Signed-off-by: Hongbin Liu <hongbinl@nvidia.com> Co-authored-by: Hongbin Liu <hongbinl@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix adding positional embeddings in-place in transformer module (#7440) Signed-off-by: Tamerlan Tabolov <tktabolov@gmail.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix (#7478) Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add sleep (#7498) (#7499) * add sleep * add sleep onto config instead * add comment --------- Signed-off-by: Gerald Shen <geshen@nvidia.com> Co-authored-by: Gerald Shen <119401249+gshennvm@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix exp manager check for sleep (#7503) (#7504) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.accelerator=auto from None. (#7492) (#7493) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [doc] fix broken link (#7481) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Read audio as int32 to avoid flac read errors (#7477) * [TTS] Read audio as int32 to avoid flac read errors Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add comment about read failures Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS (#7409) * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS * Train 'AISHELL-3' dataset with multi-speakers Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update get_data.py update copyright header Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * Update get_data.py added a disclaimer Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add new configuration file for AISHELL3 with multispeaker of fastpitch Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * dllogger - log on rank 0 only (#7513) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix TTS FastPitch tutorial (#7494) (#7516) * Fix --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix get_dist() tensor dimension (#7506) (#7515) Signed-off-by: Jocelyn Huang <jocelynh@nvidia.com> Co-authored-by: Jocelyn <jocelynh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: specify trainer.strategy=auto when devices=1 (#7509) (#7512) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix (#7511) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Fix FastPitch data prep tutorial (#7524) Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add italian tokenization (#7486) * add italian tokenization Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more ipa lexicon it Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix error deletion Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Replace None strategy with auto in tutorial notebooks (#7521) (#7527) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * unpin setuptools (#7534) (#7535) Signed-off-by: fayejf <36722593+fayejf@users.noreply.github.com> Co-authored-by: fayejf <36722593+fayejf@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * remove auto generated examples (#7510) * explicitly remove autogenerated examples for data parallel evaluation Signed-off-by: arendu <adithyare@nvidia.com> * mark autogenrated and remove it for test Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add the `strategy` argument to `MegatronGPTModel.generate()` (#7264) It is passed as an explicit argument rather than through `**strategy_args` so as to ensure someone cannot accidentally pass other arguments that would end up being ignored. It is a keyword-only argument to ensure that if in the future we want to update the signature to `**strategy_args`, we can do it without breaking code. Signed-off-by: Olivier Delalleau <507137+odelalleau@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix PTL2.0 related ASR bugs in r1.21.0: Val metrics logging, None dataloader issue (#7531) (#7533) * fix none dataloader issue ptl2 * ptl2.0 logging fixes for rnnt_models --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * gpus -> devices (#7542) (#7545) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Update FFMPEG version to fix issue with torchaudio (#7551) (#7553) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * PEFT GPT & T5 Refactor (#7308) * initial implementation of add_adapters API * correct type hint * Add config in add_adapters for save and load (@author bobchen) * Remove AdapterConfig to avoid import error * Add AdaterConfig back and move adaptermixin to sft model * Add NLPSaveRestoreConnector as default in NLPModel.restore_from * Add restore_from_nemo_with_adapter and test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * rename t5 file and classes to be consistent with GPT * add t5 sft dataset * add support for single-file format with T5SFTDataset * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Various small changes to make T5 SFT work like GPT SFT * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add adapter evaluation test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add MultiAdaterConfig for ia3 and fix builder issue * Make ptuning for T5SFTModel work using mixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add IA3_Adapter for AdapterName * Add adapter name for ptuning and attention adapter * Make test script GPT/T5 agnostic * Add layer selection feature * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Integrate adapter name and config * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update gpt peft tuning script to new API * add t5 peft tuning script with new API * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix IA3 layer selection issue * Override state_dict on SFT model instead of mixin * Add load adapter by adapter config * move peft config map away from example script * auto get config from nemo adapter * Move PEFTConfig to new file * fix ckpt save/load for t5 * name change: add_adapters -> add_adapter * variable name change * update t5 script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix t5 issues * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add weight tying * update gpt tuning script * PEFT-API proposal * Fix according to comments * update tuning scripts * move merge_cfg_with to mixin class since it applies to both gpt and t5 and requires the model class for restore * Add mcore_gpt support for NLPAdapterMixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix typo * variable name change to distinguish "peft" and "adapter" * override `load_adapters` to support `add_adapter` name change * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update tuning and eval script for adapter save/load * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add Ptuning on first stage only * add lora tutorial for review * Fix layer selection for mcore * add landing page * fix resume training Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add mcore condition in sharded_state_dict to make sft work * Update lora_tutorial.md First edit of this file for PEFT documentation for NeMO Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * rename Adapter to AttentionAdapter to avoid confusion in doc * Change load_adapters to load .nemo * add quick start guide * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add load_adapters with .ckpt * Remove setup_complete changes in load_adapters * update landing page * remove typo * Updated quick_start.md per Chen Cui Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * Add inference config merger and tutorial * Add doc string for NLPAdapterModelMixin and deprecated warning on MegatronGPTPEFTModel * add supported_methods.md and update other documentations * Update supported_methods.md minor updates. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Update landing_page.md minor update. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Modify doc string for NLPAdapterModelMixin * Add doc string add_adapters in NLPAdapterModelMixin * rename canonical adapters * remove mcore hard dependency * [PATCH] move microbatch calculator to nemo from apex * remove apex dependency in gpt and t5 sft models * remove apex dependency in gpt model * render doc strings * fix * Add missing virtual_tokens on ptuning * fix docstrings * update gpt-style model coverage in docs * update docstring * Remove pdb * add lightning_fabric to make docstring rendering work * Add Ptuning missing key * try docstring rendering * Fix ptuning issue * update gpt t5 peft tuning and eval scripts * typos * update eval config * fix bug relating to apex dependency removal * typo * make predict step behave the same as test step * make lora tutorial work in notebook * cosmetics * update yaml scripts * mcore_gpt attribute optional * typo * update eval scripts and fix T5 eval bugs * add NLPDDPStrategyNotebook and trainer builder logic to use it * update lora notebook to use new trainer builder * fix microbatch calculator bug for inference after training * Convert markdown files to RST and incorporate with doc * typo * revise language * remove extra cell * remove unnecessary inheritance * remove old tests * move layer selection default so logging messages make sense * remove `save_adapters` as adapter weights are saved automatically during training * initialize weights from a checkpoint instead of randomly * multiple fields can form a context (#7147) * list of context fields and flexible prompt template Signed-off-by: arendu <adithya.r@gmail.com> * list of fields for context Signed-off-by: arendu <adithya.r@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add multiple truncation fields and middle truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Compatible to old ckpt Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix tokenize detokenize issue Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove detokenization, add truncation augmentation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Resolve comments Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove unused import Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert eos Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add tokenizer space_sensitive attribute Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix error Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix erorr and use re Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Change assert logic Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Follow adi suggestion Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove merge function Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add example and comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove context_key and add comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove random truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix template none Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> * revert config changes * remove accidental breakpoint * support TP>1 loading * infer adapter type from checkpoint in during eval * breakup add adapter * enable interpolation of train_ds and validation_ds * update metric calc script to conform to single-file eval format * remove extraneous print * update lora notebook for updated merge_inference_cfg * Update nlp_adapter_mixins.py variable name change Signed-off-by: Chen Cui <chcui@nvidia.com> * turn off grad scaler for PP to match old scripts * remove PEFTSaveRestoreConnector since functionality all covered by the new mixin class * remove resume_from_checkpoint check since covered in #7335 * revert changes made in eval config interpolation * more interpolation * typo * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * remove dup line Signed-off-by: Chen Cui <chcui@nvidia.com> * code style warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix config mistake Signed-off-by: Chen Cui <chcui@nvidia.com> * add copyright header Signed-off-by: Chen Cui <chcui@nvidia.com> * fix code check warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert changes to remove apex dependency (mixed apex+nemo microbatch calculator broke some CI tests) Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * consolidate peft and sft scripts Signed-off-by: Chen Cui <chcui@nvidia.com> * update CI tests Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * notebook branch points to main to prepare for merge Signed-off-by: Chen Cui <chcui@nvidia.com> * fix gpt and t5 validation with any metric other than loss Signed-off-by: Chen Cui <chcui@nvidia.com> * support pre-extracted checkpoints Signed-off-by: Chen Cui <chcui@nvidia.com> --------- Signed-off-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Chen Cui <chcui@nvidia.com> Co-authored-by: Chen Cui <chcui@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Marc Romeyn <marcromeyn@gmail.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Yuanzhe Dong <yudong@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix a typo (#7496) Signed-off-by: BestJuly <chntaoli@163.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] remove curly braces from ${BRANCH} in jupyer notebook cell. (#7554) (#7560) * remove curly braces. * remove installation of pynini. --------- Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add youtube embed url (#7570) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Remap speakers to continuous range of speaker_id for dataset AISHELL3 (#7536) * Remap speakers to continuous range of speaker_id for dataset AISHELL3 * Add new key/value pair to record raw speaker for AISHELL3 dataset Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix validation_step_outputs initialization for multi-dataloader (#7546) (#7572) * added correct validation_step_outputs initialization for mutli-dataloader * changed kernel for display * Update logic for validation and test step outputs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert multidataloader changes in multilang ASR notebook --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append output of val step to self.validation_step_outputs (#7530) (#7532) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] fixed trainer's accelerator and strategy. (#7569) (#7574) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append val/test output to instance variable in EncDecSpeakerLabelModel (#7562) (#7573) * Append val/test output to the instance variable in EncDecSpeakerLabelModel * Handle test case in evaluation_step * Replace type with isinstance --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix CustomProgressBar for resume (#7427) (#7522) * Fix CustomProgress Bar for resume and multiple epochs * Edit num_training_batches * Use max_steps as total for progress bar for resume * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix typos in nfa and speech enhancement tutorials (#7580) (#7583) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add strategy as ddp_find_unused_parameters_true for glue_benchmark.py (#7454) (#7461) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * update strategy (#7577) (#7578) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix typos (#7581) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Change hifigan finetune strategy to ddp_find_unused_parameters_true (#7579) (#7584) * Change strategy to auto --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [BugFix] Add missing quotes for auto strategy in tutorial notebooks (#7541) (#7548) * Add missing quotes for auto strategy * Revert trainer.gpus to trainer.devices in Self_Supervised_Pre_Training.ipynb --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add build os key (#7596) (#7599) * add build os key * add tools * update to stable version --------- Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * StarCoder SFT test + bump PyT NGC image to 23.09 (#7540) * Add SFT StarCoder test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Remove _modify_config call as it is covered in load_from_nemo just below Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Test with pyt:23.09 container Signed-off-by: Jan Lasek <janek.lasek@gmail.com> --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * defaults changed (#7600) * defaults changed Signed-off-by: arendu <adithyare@nvidia.com> * typo Signed-off-by: arendu <adithyare@nvidia.com> * update Signed-off-by: arendu <adithyare@nvidia.com> --------- Signed-off-by: arendu <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add ItalianPhonemesTokenizer (#7587) * add ItalianPhonemesTokenizer Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix Italian phonemes Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * best ckpt fix (#7564) (#7588) Signed-off-by: dimapihtar <dpihtar@gmail.com> Co-authored-by: Dmytro Pykhtar <37850217+dimapihtar@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add files via upload (#7598) specifies the branch Signed-off-by: George <37293288+Jorjeous@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix validation in G2PModel and ThutmoseTaggerModel (#7597) (#7606) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Bound transformers version in requirements (#7620) Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix llama2 70b lora tuning bug (#7622) * fix llama2 70b lora tuning bug Signed-off-by: Chen Cui <chcui@nvidia.com> * Update peft_config.py brackets Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> --------- Signed-off-by: Chen Cui <chcui@nvidia.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix import error no module name model_utils (#7629) Signed-off-by: Mehadi Hasan Menon <mehadihasan80@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add fc large ls models (#7641) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao Koluguri <nithinraok> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.gpus, trainer.strategy, trainer.accelerator (#7621) (#7642) * [TTS] bugfix for Tacotron2 tutorial due to PTL 2.0 * trainer.gpus -> trainer.devices * fixed related tutorial bugs --------- Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix ssl models ptl monitor val through logging (#7608) (#7614) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Co-authored-by: Eric Harper <complex451@gmail.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix metrics for SE tutorial (#7604) (#7612) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Co-authored-by: anteju <108555623+anteju@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add ddp_find_unused_parameters=True and change accelerator to auto (#7623) (#7644) * Add ddp_find_unused_parameters=True and change acclerator to auto * Add ddp_find_unused_parameters True for normalization_as_tagging_train.py --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix py3.11 dataclasses issue (#7616) * Fix py3.11 dataclasses issue (#7582) * Update ASR configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Update TTS configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Guard MeCab and Ipadic Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix scripts Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Update name to ConfidenceMethodConfig Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Fix changes to confidence measure Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix issues with Dockerfile (#7650) (#7652) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] RNN-T greedy decoding max_frames fix for alignment and confidence (#7635) * decoding and test fix Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] Fix type error in jasper (#7636) (#7653) Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: Ryan Langman <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Add STFT and SI-SDR loss to audio codec recipe (#7468) * [TTS] Add STFT and SI-SDR loss to audio codec recipe Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix STFT resolution Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix training metric logging Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add docstring to mel and stft losses Signed-off-by: Ryan <rlangman@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add outline of asr quickstart info to asr/intro.rst Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add CLI, LM and real-time transcription sections Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Create per.py (#7538) * Move model precision copy (#7336) * move cfg precision set to megatron base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * remove copy from other models Signed-off-by: Maanu Grover <maanug@nvidia.com> * modify attribute not arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix gpt model test for ptl 2.0 Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename function and add docstring Signed-off-by: Maanu Grover <maanug@nvidia.com> * replace precision to dtype conditionals with func call Signed-off-by: Maanu Grover <maanug@nvidia.com> * unnecessary function and cfg reset Signed-off-by: Maanu Grover <maanug@nvidia.com> * set default value Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix precision lookup in a few more places Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename mapping function Signed-off-by: Maanu Grover <maanug@nvidia.com> * ununsed import Signed-off-by: Maanu Grover <maanug@nvidia.com> * save torch datatype to model Signed-off-by: Maanu Grover <maanug@nvidia.com> * set weights precision wrt amp o2 Signed-off-by: Maanu Grover <maanug@nvidia.com> * Revert "set weights precision wrt amp o2" This reverts commit 313a4bfe5eb69d771a6d2433898c0685836aef5c. Signed-off-by: Maanu Grover <maanug@nvidia.com> * revert half precision at inference attempt Signed-off-by: Maanu Grover <maanug@nvidia.com> * move autocast dtype to base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * move params dtype to base model, enable fp16 O2 inf Signed-off-by: Maanu Grover <maanug@nvidia.com> * unused imports Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix PEFT checkpoint loading (#7388) * Fix PEFT checkpoint loading Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use distributed optimizer support for multiple dtypes (#7359) * Update distopt wrapper with multiple dtype support Remove manual handling of separate FP32 optimizer. Signed-off-by: Tim Moon <tmoon@nvidia.com> * Use distopt support for contiguous buffers with multiple dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Fix typo Signed-off-by: Tim Moon <tmoon@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Separate distopt buckets for first GPT layer and non-overlapped params Signed-off-by: Tim Moon <tmoon@nvidia.com> * Add distopt logic for int dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit Signed-off-by: Tim Moon <tmoon@nvidia.com> * Remove unused variables Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit in README and Jenkensfile Signed-off-by: Tim Moon <tmoon@nvidia.com> * Debug Dockerfile and Jenkinsfile Signed-off-by: Tim Moon <tmoon@nvidia.com> --------- Signed-off-by: Tim Moon <tmoon@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * minor fix for llama ckpt conversion script (#7387) * minor fix for llama ckpt conversion script Signed-off-by: Jason Wang <jasonwan@nvidia.com> * Update Jenkinsfile Signed-off-by: Jason Wang <jasonwan@nvidia.com> * remove fast_swiglu configuration Signed-off-by: Jason Wang <jasonwan@nvidia.com> --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix wrong calling of librosa.get_duration() in notebook (#7376) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [PATCH] PEFT import mcore (#7393) * [PATCH] PEFT import mcore Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Create per.py Script for calculation Punctuation Error Rate and related rates (correct rate, deletions rate, etc.) Signed-off-by: Sasha Meister <117230141+ssh-meister@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Added a callback for logging initial data (#7384) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update Core Commit (#7402) * Update Core Commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * update commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use cfg attribute in bert (#7394) * use cfg attribute instead of arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * use torch_dtype in place of cfg.precision Signed-off-by: Maanu Grover <maanug@nvidia.com> * move precision copy before super constructor Signed-off-by: Maanu Grover <maanug@nvidia.com> * use trainer arg Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add support for bias conversion in Swiglu models (#7386) * Add support for bias conversion in Swiglu models Signed-off-by: smajumdar <titu1994@gmail.com> * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * Fix issue with missing tokenizer Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update save_to and restore_from for dist checkpointing (#7343) * add dist ckpt to save to, in progress Signed-off-by: eharper <eharper@nvidia.com> * move dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * clean up Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update restore from, need to figure out how to initialize distributed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * launch distrib if needed when restoring dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * when using mcore we can change tp pp on the fly Signed-off-by: eharper <eharper@nvidia.com> * add load_from_checkpoint support for dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update llama convert script to save dist .nemo Signed-off-by: eharper <eharper@nvidia.com> * fix load dist ckpt Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup TE TP groups if needed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup te tp groups if needed Signed-off-by: eharper <eharper@nvidia.com> * remove import Signed-off-by: eharper <eharper@nvidia.com> --------- Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * fix forward for with mcore=false (#7403) Signed-off-by: Jimmy Zhang <jiemingz@nvidia.com> Co-authored-by: Jimmy Zhang <jiemingz@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix logging to remove 's/it' from progress bar in Megatron models and add train_step_timing (#7374) * Add CustomProgressBar class to exp_manager and trainer callbacks Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix the progress bar to reflect total microbatch cnt Signed-off-by: Abhishree <abhishreetm@gmail.com> * Modify CustomProgressBar class 1) Modify CustomProgressBar class to update progress bar per global_step instead of per microbatch 2) Add the callback to other megatron training/finetuning files that are not using MegatronTrainerBuilder Signed-off-by: Abhishree <abhishreetm@gmail.com> * Add CustomProgressBar callback to tuning files Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Set Activation Checkpointing Defaults (#7404) * Set Activation Checkpointing Defaults Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * check for None Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * make loss mask default to false (#7407) Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add dummy userbuffer config files (#7408) Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * add missing ubconf files (#7412) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * New tutorial on Speech Data Explorer (#7405) * Added Google Colab based tutorial on Speech Data Explorer Signed-off-by: George Zelenfroynd <gzelenfroind@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update ptl training ckpt conversion script to work with dist ckpt (#7416) * update ptl convert script Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * don't break legacy Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: eharper <eharper@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Allow disabling sanity checking when num_sanity_val_steps=0 (#7413) * Allow disabling sanity checking when num_sanity_val_steps=0 Signed-off-by: Abhishree <abhishreetm@gmail.com> * Update num_sanity_val_steps to be a multiple of num_microbatches Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add comprehensive error messages (#7261) Signed-off-by: Anton Peganov <apeganov@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * check NEMO_PATH (#7418) Signed-off-by: Nikolay Karpov <karpnv@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * layer selection for ia3 (#7417) * layer selection for ia3 Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix missing pip package 'einops' (#7397) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of pyaudio in Google Colab (#7396) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update README.md: output_path --> output_manifest_filepath (#7442) Signed-off-by: Samuele Cornell <cornellsamuele@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add rope dynamic linear scaling (#7437) * Add dynamic linear scaling Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Yang Zhang <yzhang123@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix None dataloader issue in PTL2.0 (#7455) * Fix None dataloader issue in PTL2.0 Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [ASR] Confidence measure -> method renames (#7434) * measure -> method Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add steps for document of getting dataset 'SF Bilingual Speech' (#7378) * Add steps for document of getting dataset 'SF Bilingual Speech' Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update datasets.rst added a link from a tutorial demonstrating detailed data prep steps. Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * RNN-T confidence and alignment bugfix (#7381) * new frame_confidence and alignments lists are now always created after the while loop Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * tests added Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix resume from checkpoint in exp_manager (#7424) (#7426) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix checking of cuda/cpu device for inputs of Decoder (#7444) * Fix checking of cuda/cpu device for inputs of Decoder Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update tacotron2.py Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of ljspeech's get_data.py (#7430) * Fix failure of ljspeech's get_data.py Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Fix audio codec type checks (#7373) * [TTS] Fix audio codec type checks Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix audio codec tests Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Add dataset to path of logged artifacts (#7462) * [TTS] Add dataset to path of logged artifacts Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Revert axis name back to Audio Frames Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix sft dataset truncation (#7464) * Add fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Automatic Lip Reading Recognition (ALR) - ASR/CV (Visual ASR) (#7330) * striding_conv1d_k5 and dw_striding_conv1d_k5 subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * transpose conv1d inputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update subsampling.py change striding_conv1d_k5 to striding_conv1d Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * video manifest Signed-off-by: mburchi <maxime.burchi@gmail.com> * add collection classes Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-… * fix onnx (#7703) (#7704) Signed-off-by: fayejf <fayejf07@gmail.com> Co-authored-by: fayejf <36722593+fayejf@users.noreply.github.com> * move core install to /workspace (#7706) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * Fix typo in audio codec config, encoder target (#7697) Signed-off-by: Ante Jukić <ajukic@nvidia.com> * Replace strategy='dp'/None with 'auto' (#7681) (#7696) * Add strategy=auto for None and dp * Change strategy from None to auto --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> * [ASR] Multichannel mask estimator with flex number of channels (#7317) * Adding a mask estimator which can …

* Fix multi rank finetune for ASR * Actually add time * Actually add time --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com>

) * Hotfix (#7501) (#7568) Signed-off-by: Jan Baczek <jbaczek@nvidia.com> Co-authored-by: jbaczek <45043825+jbaczek@users.noreply.github.com> * Avoid duplicated checkpoint save (#7555) (#7566) Signed-off-by: Mikołaj Błaż <mblaz@nvidia.com> Co-authored-by: mikolajblaz <mikolajblaz@users.noreply.github.com> * Cache FP8 weight and transpose only at the first micro-batch in each validation and test routine (#7470) (#7483) * Cache weight and transpose only in the first batch in all training, val, and test runs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Add an option to disable manual GC in validation (#7467) (#7476) Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> * Remove PUBLICATIONS.md, point to github.io NeMo page instead (#7694) (#7695) * update publications section to point to blog website page * add hyphen * use double backquotes for code formatting --------- Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Signed-off-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> * Fix multi rank finetune for ASR (#7684) (#7699) * Fix multi rank finetune for ASR * Actually add time * Actually add time --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> * Update docs: readme, getting started, ASR intro (#7679) * [TTS] Add dataset to path of logged artifacts (#7462) * [TTS] Add dataset to path of logged artifacts Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Revert axis name back to Audio Frames Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * move install info to INSTALLATION.md Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * tidy up links Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix sft dataset truncation (#7464) * Add fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Automatic Lip Reading Recognition (ALR) - ASR/CV (Visual ASR) (#7330) * striding_conv1d_k5 and dw_striding_conv1d_k5 subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * transpose conv1d inputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update subsampling.py change striding_conv1d_k5 to striding_conv1d Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * video manifest Signed-off-by: mburchi <maxime.burchi@gmail.com> * add collection classes Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test_step_outputs Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest bug when having only audio or only videos Signed-off-by: mburchi <maxime.burchi@gmail.com> * clean references Signed-off-by: mburchi <maxime.burchi@gmail.com> * freeze unfreeze transcribe cv models Signed-off-by: mburchi <maxime.burchi@gmail.com> * correct manifest get_full_path bug Signed-off-by: mburchi <maxime.burchi@gmail.com> * update for PR Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * guard torchvision Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * _video_speech_collate_fn in cv/data/video_to_text.py Signed-off-by: mburchi <maxime.burchi@gmail.com> * add self.out = None to asr subsampling Signed-off-by: mburchi <maxime.burchi@gmail.com> * Update nemo/collections/cv/data/video_to_text_dataset.py Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> * cv -> multimodal/speech_cv branch Signed-off-by: mburchi <maxime.burchi@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: mburchi <maxime.burchi@gmail.com> Signed-off-by: Maxime Burchi <60737204+burchim@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Igor Gitman <igor.a.gitman@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * HF StarCoder to NeMo conversion script (#7421) * Script to convert HF StarCoder checkpoint to NeMo Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * StarCoder conversion test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Fix test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Catch up with save_to changes Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Don't abbreviate args for clarity Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Configurable precision: BF16 vs FP32 Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix bug when loading dist ckpt in peft (#7452) Signed-off-by: Hongbin Liu <hongbinl@nvidia.com> Co-authored-by: Hongbin Liu <hongbinl@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix adding positional embeddings in-place in transformer module (#7440) Signed-off-by: Tamerlan Tabolov <tktabolov@gmail.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix (#7478) Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add sleep (#7498) (#7499) * add sleep * add sleep onto config instead * add comment --------- Signed-off-by: Gerald Shen <geshen@nvidia.com> Co-authored-by: Gerald Shen <119401249+gshennvm@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix exp manager check for sleep (#7503) (#7504) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.accelerator=auto from None. (#7492) (#7493) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [doc] fix broken link (#7481) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Read audio as int32 to avoid flac read errors (#7477) * [TTS] Read audio as int32 to avoid flac read errors Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add comment about read failures Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS (#7409) * Add dataset 'AISHELL-3' from OpenSLR for training mandarin TTS * Train 'AISHELL-3' dataset with multi-speakers Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update get_data.py update copyright header Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * Update get_data.py added a disclaimer Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add new configuration file for AISHELL3 with multispeaker of fastpitch Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * dllogger - log on rank 0 only (#7513) Signed-off-by: Stas Bekman <stas00@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix TTS FastPitch tutorial (#7494) (#7516) * Fix --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix get_dist() tensor dimension (#7506) (#7515) Signed-off-by: Jocelyn Huang <jocelynh@nvidia.com> Co-authored-by: Jocelyn <jocelynh@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: specify trainer.strategy=auto when devices=1 (#7509) (#7512) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix (#7511) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Fix FastPitch data prep tutorial (#7524) Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add italian tokenization (#7486) * add italian tokenization Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more ipa lexicon it Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix error deletion Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Replace None strategy with auto in tutorial notebooks (#7521) (#7527) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * unpin setuptools (#7534) (#7535) Signed-off-by: fayejf <36722593+fayejf@users.noreply.github.com> Co-authored-by: fayejf <36722593+fayejf@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * remove auto generated examples (#7510) * explicitly remove autogenerated examples for data parallel evaluation Signed-off-by: arendu <adithyare@nvidia.com> * mark autogenrated and remove it for test Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add the `strategy` argument to `MegatronGPTModel.generate()` (#7264) It is passed as an explicit argument rather than through `**strategy_args` so as to ensure someone cannot accidentally pass other arguments that would end up being ignored. It is a keyword-only argument to ensure that if in the future we want to update the signature to `**strategy_args`, we can do it without breaking code. Signed-off-by: Olivier Delalleau <507137+odelalleau@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix PTL2.0 related ASR bugs in r1.21.0: Val metrics logging, None dataloader issue (#7531) (#7533) * fix none dataloader issue ptl2 * ptl2.0 logging fixes for rnnt_models --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * gpus -> devices (#7542) (#7545) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Update FFMPEG version to fix issue with torchaudio (#7551) (#7553) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * PEFT GPT & T5 Refactor (#7308) * initial implementation of add_adapters API * correct type hint * Add config in add_adapters for save and load (@author bobchen) * Remove AdapterConfig to avoid import error * Add AdaterConfig back and move adaptermixin to sft model * Add NLPSaveRestoreConnector as default in NLPModel.restore_from * Add restore_from_nemo_with_adapter and test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * rename t5 file and classes to be consistent with GPT * add t5 sft dataset * add support for single-file format with T5SFTDataset * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Various small changes to make T5 SFT work like GPT SFT * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add adapter evaluation test script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add MultiAdaterConfig for ia3 and fix builder issue * Make ptuning for T5SFTModel work using mixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add IA3_Adapter for AdapterName * Add adapter name for ptuning and attention adapter * Make test script GPT/T5 agnostic * Add layer selection feature * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Integrate adapter name and config * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update gpt peft tuning script to new API * add t5 peft tuning script with new API * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix IA3 layer selection issue * Override state_dict on SFT model instead of mixin * Add load adapter by adapter config * move peft config map away from example script * auto get config from nemo adapter * Move PEFTConfig to new file * fix ckpt save/load for t5 * name change: add_adapters -> add_adapter * variable name change * update t5 script * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix t5 issues * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add weight tying * update gpt tuning script * PEFT-API proposal * Fix according to comments * update tuning scripts * move merge_cfg_with to mixin class since it applies to both gpt and t5 and requires the model class for restore * Add mcore_gpt support for NLPAdapterMixin * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix typo * variable name change to distinguish "peft" and "adapter" * override `load_adapters` to support `add_adapter` name change * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update tuning and eval script for adapter save/load * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add Ptuning on first stage only * add lora tutorial for review * Fix layer selection for mcore * add landing page * fix resume training Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add mcore condition in sharded_state_dict to make sft work * Update lora_tutorial.md First edit of this file for PEFT documentation for NeMO Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * rename Adapter to AttentionAdapter to avoid confusion in doc * Change load_adapters to load .nemo * add quick start guide * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add load_adapters with .ckpt * Remove setup_complete changes in load_adapters * update landing page * remove typo * Updated quick_start.md per Chen Cui Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> * Add inference config merger and tutorial * Add doc string for NLPAdapterModelMixin and deprecated warning on MegatronGPTPEFTModel * add supported_methods.md and update other documentations * Update supported_methods.md minor updates. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Update landing_page.md minor update. Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> * Modify doc string for NLPAdapterModelMixin * Add doc string add_adapters in NLPAdapterModelMixin * rename canonical adapters * remove mcore hard dependency * [PATCH] move microbatch calculator to nemo from apex * remove apex dependency in gpt and t5 sft models * remove apex dependency in gpt model * render doc strings * fix * Add missing virtual_tokens on ptuning * fix docstrings * update gpt-style model coverage in docs * update docstring * Remove pdb * add lightning_fabric to make docstring rendering work * Add Ptuning missing key * try docstring rendering * Fix ptuning issue * update gpt t5 peft tuning and eval scripts * typos * update eval config * fix bug relating to apex dependency removal * typo * make predict step behave the same as test step * make lora tutorial work in notebook * cosmetics * update yaml scripts * mcore_gpt attribute optional * typo * update eval scripts and fix T5 eval bugs * add NLPDDPStrategyNotebook and trainer builder logic to use it * update lora notebook to use new trainer builder * fix microbatch calculator bug for inference after training * Convert markdown files to RST and incorporate with doc * typo * revise language * remove extra cell * remove unnecessary inheritance * remove old tests * move layer selection default so logging messages make sense * remove `save_adapters` as adapter weights are saved automatically during training * initialize weights from a checkpoint instead of randomly * multiple fields can form a context (#7147) * list of context fields and flexible prompt template Signed-off-by: arendu <adithya.r@gmail.com> * list of fields for context Signed-off-by: arendu <adithya.r@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add multiple truncation fields and middle truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Compatible to old ckpt Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix tokenize detokenize issue Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove detokenization, add truncation augmentation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Resolve comments Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove unused import Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert eos Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Add tokenizer space_sensitive attribute Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix error Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Fix erorr and use re Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Change assert logic Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Follow adi suggestion Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove merge function Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add example and comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove context_key and add comment Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * Remove random truncation Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix template none Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> * revert config changes * remove accidental breakpoint * support TP>1 loading * infer adapter type from checkpoint in during eval * breakup add adapter * enable interpolation of train_ds and validation_ds * update metric calc script to conform to single-file eval format * remove extraneous print * update lora notebook for updated merge_inference_cfg * Update nlp_adapter_mixins.py variable name change Signed-off-by: Chen Cui <chcui@nvidia.com> * turn off grad scaler for PP to match old scripts * remove PEFTSaveRestoreConnector since functionality all covered by the new mixin class * remove resume_from_checkpoint check since covered in #7335 * revert changes made in eval config interpolation * more interpolation * typo * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * remove dup line Signed-off-by: Chen Cui <chcui@nvidia.com> * code style warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix config mistake Signed-off-by: Chen Cui <chcui@nvidia.com> * add copyright header Signed-off-by: Chen Cui <chcui@nvidia.com> * fix code check warnings Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert changes to remove apex dependency (mixed apex+nemo microbatch calculator broke some CI tests) Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add more deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * update deprecation notices Signed-off-by: Chen Cui <chcui@nvidia.com> * consolidate peft and sft scripts Signed-off-by: Chen Cui <chcui@nvidia.com> * update CI tests Signed-off-by: Chen Cui <chcui@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * notebook branch points to main to prepare for merge Signed-off-by: Chen Cui <chcui@nvidia.com> * fix gpt and t5 validation with any metric other than loss Signed-off-by: Chen Cui <chcui@nvidia.com> * support pre-extracted checkpoints Signed-off-by: Chen Cui <chcui@nvidia.com> --------- Signed-off-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: arendu <adithya.r@gmail.com> Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Signed-off-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Chen Cui <chcui@nvidia.com> Co-authored-by: Chen Cui <chcui@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Marc Romeyn <marcromeyn@gmail.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: hkelly33 <58792115+hkelly33@users.noreply.github.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Yuanzhe Dong <yudong@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix a typo (#7496) Signed-off-by: BestJuly <chntaoli@163.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] remove curly braces from ${BRANCH} in jupyer notebook cell. (#7554) (#7560) * remove curly braces. * remove installation of pynini. --------- Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add youtube embed url (#7570) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Remap speakers to continuous range of speaker_id for dataset AISHELL3 (#7536) * Remap speakers to continuous range of speaker_id for dataset AISHELL3 * Add new key/value pair to record raw speaker for AISHELL3 dataset Signed-off-by: Robin Dong <robin.k.dong@gmail.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix validation_step_outputs initialization for multi-dataloader (#7546) (#7572) * added correct validation_step_outputs initialization for mutli-dataloader * changed kernel for display * Update logic for validation and test step outputs * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * revert multidataloader changes in multilang ASR notebook --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Kunal Dhawan <kunaldhawan97@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append output of val step to self.validation_step_outputs (#7530) (#7532) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] fixed trainer's accelerator and strategy. (#7569) (#7574) Signed-off-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <16880-xueyang@users.noreply.gitlab-master.nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Append val/test output to instance variable in EncDecSpeakerLabelModel (#7562) (#7573) * Append val/test output to the instance variable in EncDecSpeakerLabelModel * Handle test case in evaluation_step * Replace type with isinstance --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix CustomProgressBar for resume (#7427) (#7522) * Fix CustomProgress Bar for resume and multiple epochs * Edit num_training_batches * Use max_steps as total for progress bar for resume * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix typos in nfa and speech enhancement tutorials (#7580) (#7583) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> Co-authored-by: Elena Rastorgueva <80532067+erastorgueva-nv@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add strategy as ddp_find_unused_parameters_true for glue_benchmark.py (#7454) (#7461) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * update strategy (#7577) (#7578) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix typos (#7581) Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Change hifigan finetune strategy to ddp_find_unused_parameters_true (#7579) (#7584) * Change strategy to auto --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: Cheng-Ping Hsieh <37269846+hsiehjackson@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [BugFix] Add missing quotes for auto strategy in tutorial notebooks (#7541) (#7548) * Add missing quotes for auto strategy * Revert trainer.gpus to trainer.devices in Self_Supervised_Pre_Training.ipynb --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add build os key (#7596) (#7599) * add build os key * add tools * update to stable version --------- Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * StarCoder SFT test + bump PyT NGC image to 23.09 (#7540) * Add SFT StarCoder test Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Remove _modify_config call as it is covered in load_from_nemo just below Signed-off-by: Jan Lasek <janek.lasek@gmail.com> * Test with pyt:23.09 container Signed-off-by: Jan Lasek <janek.lasek@gmail.com> --------- Signed-off-by: Jan Lasek <janek.lasek@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * defaults changed (#7600) * defaults changed Signed-off-by: arendu <adithyare@nvidia.com> * typo Signed-off-by: arendu <adithyare@nvidia.com> * update Signed-off-by: arendu <adithyare@nvidia.com> --------- Signed-off-by: arendu <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add ItalianPhonemesTokenizer (#7587) * add ItalianPhonemesTokenizer Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix Italian phonemes Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * add test Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> --------- Signed-off-by: GiacomoLeoneMaria <giacomoleonemaria@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * best ckpt fix (#7564) (#7588) Signed-off-by: dimapihtar <dpihtar@gmail.com> Co-authored-by: Dmytro Pykhtar <37850217+dimapihtar@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add files via upload (#7598) specifies the branch Signed-off-by: George <37293288+Jorjeous@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix validation in G2PModel and ThutmoseTaggerModel (#7597) (#7606) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Bound transformers version in requirements (#7620) Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix llama2 70b lora tuning bug (#7622) * fix llama2 70b lora tuning bug Signed-off-by: Chen Cui <chcui@nvidia.com> * Update peft_config.py brackets Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> --------- Signed-off-by: Chen Cui <chcui@nvidia.com> Signed-off-by: Adi Renduchintala <adithyare@nvidia.com> Co-authored-by: Adi Renduchintala <adithyare@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix import error no module name model_utils (#7629) Signed-off-by: Mehadi Hasan Menon <mehadihasan80@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add fc large ls models (#7641) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao Koluguri <nithinraok> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * bugfix: trainer.gpus, trainer.strategy, trainer.accelerator (#7621) (#7642) * [TTS] bugfix for Tacotron2 tutorial due to PTL 2.0 * trainer.gpus -> trainer.devices * fixed related tutorial bugs --------- Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * fix ssl models ptl monitor val through logging (#7608) (#7614) Signed-off-by: Nithin Rao Koluguri <nithinraok> Co-authored-by: Nithin Rao <nithinrao.koluguri@gmail.com> Co-authored-by: Eric Harper <complex451@gmail.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix metrics for SE tutorial (#7604) (#7612) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Co-authored-by: anteju <108555623+anteju@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Add ddp_find_unused_parameters=True and change accelerator to auto (#7623) (#7644) * Add ddp_find_unused_parameters=True and change acclerator to auto * Add ddp_find_unused_parameters True for normalization_as_tagging_train.py --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix py3.11 dataclasses issue (#7616) * Fix py3.11 dataclasses issue (#7582) * Update ASR configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Update TTS configs to support Python 3.11 Signed-off-by: smajumdar <titu1994@gmail.com> * Guard MeCab and Ipadic Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix remaining ASR dataclasses Signed-off-by: smajumdar <titu1994@gmail.com> * Fix scripts Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Update name to ConfidenceMethodConfig Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain (#7576) (#7586) * Broadcast loss only when using pipeline parallelism and within the pipeline parallel domain * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Safeguard nemo_text_processing installation on ARM (#7485) * safeguard nemo_text_processing installing Signed-off-by: Jason <jasoli@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update check Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> * Fix changes to confidence measure Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Sangkug Lym <slym@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Fix issues with Dockerfile (#7650) (#7652) Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] RNN-T greedy decoding max_frames fix for alignment and confidence (#7635) * decoding and test fix Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [ASR] Fix type error in jasper (#7636) (#7653) Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: Ryan Langman <rlangman@nvidia.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * [TTS] Add STFT and SI-SDR loss to audio codec recipe (#7468) * [TTS] Add STFT and SI-SDR loss to audio codec recipe Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix STFT resolution Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix training metric logging Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Add docstring to mel and stft losses Signed-off-by: Ryan <rlangman@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Ryan <rlangman@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add outline of asr quickstart info to asr/intro.rst Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * add CLI, LM and real-time transcription sections Signed-off-by: Elena Rastorgueva <erastorgueva@nvidia.com> * Create per.py (#7538) * Move model precision copy (#7336) * move cfg precision set to megatron base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * remove copy from other models Signed-off-by: Maanu Grover <maanug@nvidia.com> * modify attribute not arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix gpt model test for ptl 2.0 Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename function and add docstring Signed-off-by: Maanu Grover <maanug@nvidia.com> * replace precision to dtype conditionals with func call Signed-off-by: Maanu Grover <maanug@nvidia.com> * unnecessary function and cfg reset Signed-off-by: Maanu Grover <maanug@nvidia.com> * set default value Signed-off-by: Maanu Grover <maanug@nvidia.com> * fix precision lookup in a few more places Signed-off-by: Maanu Grover <maanug@nvidia.com> * rename mapping function Signed-off-by: Maanu Grover <maanug@nvidia.com> * ununsed import Signed-off-by: Maanu Grover <maanug@nvidia.com> * save torch datatype to model Signed-off-by: Maanu Grover <maanug@nvidia.com> * set weights precision wrt amp o2 Signed-off-by: Maanu Grover <maanug@nvidia.com> * Revert "set weights precision wrt amp o2" This reverts commit 313a4bfe5eb69d771a6d2433898c0685836aef5c. Signed-off-by: Maanu Grover <maanug@nvidia.com> * revert half precision at inference attempt Signed-off-by: Maanu Grover <maanug@nvidia.com> * move autocast dtype to base model Signed-off-by: Maanu Grover <maanug@nvidia.com> * move params dtype to base model, enable fp16 O2 inf Signed-off-by: Maanu Grover <maanug@nvidia.com> * unused imports Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix PEFT checkpoint loading (#7388) * Fix PEFT checkpoint loading Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use distributed optimizer support for multiple dtypes (#7359) * Update distopt wrapper with multiple dtype support Remove manual handling of separate FP32 optimizer. Signed-off-by: Tim Moon <tmoon@nvidia.com> * Use distopt support for contiguous buffers with multiple dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Fix typo Signed-off-by: Tim Moon <tmoon@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Separate distopt buckets for first GPT layer and non-overlapped params Signed-off-by: Tim Moon <tmoon@nvidia.com> * Add distopt logic for int dtypes Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit Signed-off-by: Tim Moon <tmoon@nvidia.com> * Remove unused variables Signed-off-by: Tim Moon <tmoon@nvidia.com> * Update Apex commit in README and Jenkensfile Signed-off-by: Tim Moon <tmoon@nvidia.com> * Debug Dockerfile and Jenkinsfile Signed-off-by: Tim Moon <tmoon@nvidia.com> --------- Signed-off-by: Tim Moon <tmoon@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * minor fix for llama ckpt conversion script (#7387) * minor fix for llama ckpt conversion script Signed-off-by: Jason Wang <jasonwan@nvidia.com> * Update Jenkinsfile Signed-off-by: Jason Wang <jasonwan@nvidia.com> * remove fast_swiglu configuration Signed-off-by: Jason Wang <jasonwan@nvidia.com> --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix wrong calling of librosa.get_duration() in notebook (#7376) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: Somshubra Majumdar <titu1994@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [PATCH] PEFT import mcore (#7393) * [PATCH] PEFT import mcore Signed-off-by: Jason Wang <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Jason Wang <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Create per.py Script for calculation Punctuation Error Rate and related rates (correct rate, deletions rate, etc.) Signed-off-by: Sasha Meister <117230141+ssh-meister@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Added a callback for logging initial data (#7384) Signed-off-by: Ante Jukić <ajukic@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update Core Commit (#7402) * Update Core Commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * update commit Signed-off-by: Abhinav Khattar <aklife97@gmail.com> --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Use cfg attribute in bert (#7394) * use cfg attribute instead of arg Signed-off-by: Maanu Grover <maanug@nvidia.com> * use torch_dtype in place of cfg.precision Signed-off-by: Maanu Grover <maanug@nvidia.com> * move precision copy before super constructor Signed-off-by: Maanu Grover <maanug@nvidia.com> * use trainer arg Signed-off-by: Maanu Grover <maanug@nvidia.com> --------- Signed-off-by: Maanu Grover <maanug@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add support for bias conversion in Swiglu models (#7386) * Add support for bias conversion in Swiglu models Signed-off-by: smajumdar <titu1994@gmail.com> * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add support for auto extracting tokenizer model Signed-off-by: smajumdar <titu1994@gmail.com> * Fix issue with missing tokenizer Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * Refactor Signed-off-by: smajumdar <titu1994@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: smajumdar <titu1994@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update save_to and restore_from for dist checkpointing (#7343) * add dist ckpt to save to, in progress Signed-off-by: eharper <eharper@nvidia.com> * move dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * clean up Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update restore from, need to figure out how to initialize distributed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * launch distrib if needed when restoring dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * when using mcore we can change tp pp on the fly Signed-off-by: eharper <eharper@nvidia.com> * add load_from_checkpoint support for dist ckpt Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update llama convert script to save dist .nemo Signed-off-by: eharper <eharper@nvidia.com> * fix load dist ckpt Signed-off-by: jasonwan <jasonwan@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup TE TP groups if needed Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * setup te tp groups if needed Signed-off-by: eharper <eharper@nvidia.com> * remove import Signed-off-by: eharper <eharper@nvidia.com> --------- Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: jasonwan <jasonwan@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: jasonwan <jasonwan@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * fix forward for with mcore=false (#7403) Signed-off-by: Jimmy Zhang <jiemingz@nvidia.com> Co-authored-by: Jimmy Zhang <jiemingz@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix logging to remove 's/it' from progress bar in Megatron models and add train_step_timing (#7374) * Add CustomProgressBar class to exp_manager and trainer callbacks Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix the progress bar to reflect total microbatch cnt Signed-off-by: Abhishree <abhishreetm@gmail.com> * Modify CustomProgressBar class 1) Modify CustomProgressBar class to update progress bar per global_step instead of per microbatch 2) Add the callback to other megatron training/finetuning files that are not using MegatronTrainerBuilder Signed-off-by: Abhishree <abhishreetm@gmail.com> * Add CustomProgressBar callback to tuning files Signed-off-by: Abhishree <abhishreetm@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Set Activation Checkpointing Defaults (#7404) * Set Activation Checkpointing Defaults Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * check for None Signed-off-by: Abhinav Khattar <aklife97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * make loss mask default to false (#7407) Signed-off-by: eharper <eharper@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add dummy userbuffer config files (#7408) Signed-off-by: Sangkug Lym <slym@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * add missing ubconf files (#7412) Signed-off-by: Abhinav Khattar <aklife97@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * New tutorial on Speech Data Explorer (#7405) * Added Google Colab based tutorial on Speech Data Explorer Signed-off-by: George Zelenfroynd <gzelenfroind@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update ptl training ckpt conversion script to work with dist ckpt (#7416) * update ptl convert script Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * don't break legacy Signed-off-by: eharper <eharper@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: eharper <eharper@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Allow disabling sanity checking when num_sanity_val_steps=0 (#7413) * Allow disabling sanity checking when num_sanity_val_steps=0 Signed-off-by: Abhishree <abhishreetm@gmail.com> * Update num_sanity_val_steps to be a multiple of num_microbatches Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Abhishree <abhishreetm@gmail.com> Signed-off-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add comprehensive error messages (#7261) Signed-off-by: Anton Peganov <apeganov@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * check NEMO_PATH (#7418) Signed-off-by: Nikolay Karpov <karpnv@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * layer selection for ia3 (#7417) * layer selection for ia3 Signed-off-by: arendu <adithyare@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: arendu <adithyare@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix missing pip package 'einops' (#7397) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of pyaudio in Google Colab (#7396) Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Update README.md: output_path --> output_manifest_filepath (#7442) Signed-off-by: Samuele Cornell <cornellsamuele@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add rope dynamic linear scaling (#7437) * Add dynamic linear scaling Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix bug Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> --------- Signed-off-by: Cheng-Ping Hsieh <chsieh@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Yang Zhang <yzhang123@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix None dataloader issue in PTL2.0 (#7455) * Fix None dataloader issue in PTL2.0 Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * updating values of self._validation_dl and self._test_dl as well Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: KunalDhawan <kunaldhawan97@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [ASR] Confidence measure -> method renames (#7434) * measure -> method Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Add steps for document of getting dataset 'SF Bilingual Speech' (#7378) * Add steps for document of getting dataset 'SF Bilingual Speech' Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update datasets.rst added a link from a tutorial demonstrating detailed data prep steps. Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Co-authored-by: Xuesong Yang <1646669+XuesongYang@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * RNN-T confidence and alignment bugfix (#7381) * new frame_confidence and alignments lists are now always created after the while loop Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> * tests added Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> --------- Signed-off-by: Aleksandr Laptev <alaptev@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix resume from checkpoint in exp_manager (#7424) (#7426) Signed-off-by: Abhishree <abhishreetm@gmail.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Eric Harper <complex451@gmail.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix checking of cuda/cpu device for inputs of Decoder (#7444) * Fix checking of cuda/cpu device for inputs of Decoder Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * Update tacotron2.py Signed-off-by: Jason <jasoli@nvidia.com> --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Signed-off-by: Jason <jasoli@nvidia.com> Co-authored-by: Jason <jasoli@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * Fix failure of ljspeech's get_data.py (#7430) * Fix failure of ljspeech's get_data.py Signed-off-by: Robin Dong <robin.k.dong@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Robin Dong <robin.k.dong@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * [TTS] Fix audio codec type checks (#7373) * [TTS] Fix audio codec type checks Signed-off-by: Ryan <rlangman@nvidia.com> * [TTS] Fix audio codec tests Signed-off-by: Ryan <rlangman@nvidia.com> --------- Signed-off-by: Ryan <rlangman@nvidia.com> Signed-off-by: Sasha Meister <sasha.meister.work@gmail.com> * […

titu1994 requested a review from nithinraok October 11, 2023 02:34

github-actions bot added the ASR label Oct 11, 2023

nithinraok previously approved these changes Oct 11, 2023

View reviewed changes

titu1994 added 3 commits October 11, 2023 11:20

Fix multi rank finetune for ASR

a264752

Signed-off-by: smajumdar <titu1994@gmail.com>

Actually add time

e7e9db9

Signed-off-by: smajumdar <titu1994@gmail.com>

Actually add time

0bd7424

Signed-off-by: smajumdar <titu1994@gmail.com>

titu1994 dismissed nithinraok’s stale review via 0bd7424 October 11, 2023 18:21

titu1994 force-pushed the fix_finetune branch from c031a44 to 0bd7424 Compare October 11, 2023 18:21

titu1994 merged commit 4ac544b into r1.21.0 Oct 11, 2023
10 checks passed

titu1994 deleted the fix_finetune branch October 11, 2023 21:57

titu1994 restored the fix_finetune branch October 12, 2023 00:17

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix multi rank finetune for ASR #7684

Fix multi rank finetune for ASR #7684

titu1994 commented Oct 11, 2023

nithinraok Oct 11, 2023

titu1994 Oct 11, 2023

Fix multi rank finetune for ASR #7684

Fix multi rank finetune for ASR #7684

Conversation

titu1994 commented Oct 11, 2023

What does this PR do ?

Changelog

Before your PR is "Ready for review"

nithinraok Oct 11, 2023

Choose a reason for hiding this comment

titu1994 Oct 11, 2023

Choose a reason for hiding this comment