Dataset v3.0.0 deprecates tasks and cause CI failures

[vidyasiv]

System Info

Main: Latest
Docker: 1.18.0/1.17.0

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

Test run:

pytest -v -s tests/test_examples.py -k test_run_speech_recognition_ctc_wav2vec2-large-lv60_multi_card

Output:

[rank2]:     from datasets.tasks import AutomaticSpeechRecognition
[rank2]: ModuleNotFoundError: No module named 'datasets.tasks'

This is from https://huggingface.co/datasets/regisss/librispeech_asr_for_optimum_habana_ci/blob/main/librispeech_asr_for_optimum_habana_ci.py#L23

Reason is datasets v3.0.0 removes tasks per release notes and PR
Possibly we need to downgrade datasets or find an alternative implementation that works with v3.0.0

Expected behavior

Test should pass

[vidyasiv]

@regisss , please review PR: https://huggingface.co/datasets/regisss/librispeech_asr_for_optimum_habana_ci/discussions/1/

[regisss]

It should work now with https://huggingface.co/datasets/regisss/librispeech_asr_for_optimum_habana_ci/commit/e15120eeee593aff1ed38f987d99ac30cd17a168.
It's the same change as for the original dataset: https://huggingface.co/datasets/openslr/librispeech_asr/commit/85efb3e7036e87a84b6ecf630b7d168b3c2c2696

Does the test pass on your side?

[regisss]

@regisss , please review PR: https://huggingface.co/datasets/regisss/librispeech_asr_for_optimum_habana_ci/discussions/1/

Arf sorry I didn't see this comment :/

[vidyasiv]

@regisss , please review PR: https://huggingface.co/datasets/regisss/librispeech_asr_for_optimum_habana_ci/discussions/1/

Arf sorry I didn't see this comment :/

No problem, I am getting past the failure now thanks.