Remove precision args in trainer due to PTL update (NVIDIA#8908)

* Fix precision args in trainer due to PTL update Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * roll back one change Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com> --------- Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Abhishree Thittenamane <47577437+athitten@users.noreply.github.com> Co-authored-by: Pablo Garay <palenq@gmail.com>
alxzhang-amazon · Apr 26, 2024 · 31e309e · 31e309e
1 parent 19bdda9
commit 31e309e
Show file tree

Hide file tree

Showing 7 changed files with 9 additions and 13 deletions.
diff --git a/examples/multimodal/multimodal_llm/neva/convert_hf_llava_to_neva.py b/examples/multimodal/multimodal_llm/neva/convert_hf_llava_to_neva.py
@@ -205,7 +205,7 @@ def convert(args):
     nemo_config.precision = precision
     print(f"nemo_config: {nemo_config}")
 
-    trainer = Trainer(plugins=plugins, accelerator='cpu', precision=precision, strategy=NLPDDPStrategy())
+    trainer = Trainer(plugins=plugins, accelerator='cpu', strategy=NLPDDPStrategy())
 
     hidden_size = hf_config["hidden_size"]
     head_num = hf_config["num_attention_heads"]

diff --git a/examples/nlp/language_modeling/megatron_change_num_partitions.py b/examples/nlp/language_modeling/megatron_change_num_partitions.py
@@ -938,7 +938,7 @@ def main():
         # Set precision None after precision plugins are created as PTL >= 2.1 does not allow both
         # precision plugins and precision to exist
         precision = None
-    trainer = Trainer(plugins=plugins, devices=1, strategy=NLPDDPStrategy(), accelerator="cpu", precision=precision)
+    trainer = Trainer(plugins=plugins, devices=1, strategy=NLPDDPStrategy(), accelerator="cpu")
 
     if tp_size < 0 or pp_size < 0:
         logging.info(f"Loading model config from {args.model_file} to get TP and PP size")
@@ -1205,9 +1205,7 @@ def main():
         if vp_size > 1:
             set_virtual_parallel_rank_safely(None)
 
-        trainer = Trainer(
-            plugins=plugins, devices=1, strategy=NLPDDPStrategy(), accelerator="cpu", precision=precision
-        )
+        trainer = Trainer(plugins=plugins, devices=1, strategy=NLPDDPStrategy(), accelerator="cpu")
 
         with open_dict(model.cfg):
             if args.tokenizer_model_path is not None:
@@ -1413,9 +1411,7 @@ def main():
                 app_state.pipeline_model_parallel_size * app_state.tensor_model_parallel_size
             )
 
-            trainer = Trainer(
-                plugins=plugins, devices=1, strategy=NLPDDPStrategy(), accelerator="cpu", precision=precision
-            )
+            trainer = Trainer(plugins=plugins, devices=1, strategy=NLPDDPStrategy(), accelerator="cpu")
             if args.tokenizer_model_path is not None:
                 with open_dict(model.cfg):
                     model.cfg.tokenizer.model = args.tokenizer_model_path

diff --git a/scripts/checkpoint_converters/convert_baichuan2_hf_to_nemo.py b/scripts/checkpoint_converters/convert_baichuan2_hf_to_nemo.py
@@ -175,7 +175,7 @@ def convert(args):
     nemo_config.precision = precision
     print(f"nemo_config: {nemo_config}")
 
-    trainer = Trainer(plugins=plugins, accelerator='cpu', precision=precision, strategy=NLPDDPStrategy())
+    trainer = Trainer(plugins=plugins, accelerator='cpu', strategy=NLPDDPStrategy())
 
     hidden_size = hf_config["hidden_size"]
     head_num = hf_config["num_attention_heads"]

diff --git a/scripts/checkpoint_converters/convert_chatglm_hf_to_nemo.py b/scripts/checkpoint_converters/convert_chatglm_hf_to_nemo.py
@@ -142,7 +142,7 @@ def convert(args):
 
     nemo_config.precision = precision
 
-    trainer = Trainer(plugins=plugins, accelerator='cpu', precision=precision, strategy=NLPDDPStrategy())
+    trainer = Trainer(plugins=plugins, accelerator='cpu', strategy=NLPDDPStrategy())
 
     hidden_size = hf_config["hidden_size"]
     head_num = hf_config["num_attention_heads"]

diff --git a/scripts/checkpoint_converters/convert_mistral_7b_hf_to_nemo.py b/scripts/checkpoint_converters/convert_mistral_7b_hf_to_nemo.py
@@ -193,7 +193,7 @@ def convert(args):
     nemo_config.precision = precision
     logging.info(f"nemo_config: {nemo_config}")
 
-    trainer = Trainer(plugins=plugins, accelerator='cpu', precision=precision, strategy=NLPDDPStrategy())
+    trainer = Trainer(plugins=plugins, accelerator='cpu', strategy=NLPDDPStrategy())
 
     hidden_size = nemo_config.hidden_size
     head_num = nemo_config.num_attention_heads

diff --git a/scripts/checkpoint_converters/convert_mixtral_hf_to_nemo.py b/scripts/checkpoint_converters/convert_mixtral_hf_to_nemo.py
@@ -194,7 +194,7 @@ def convert(args):
     nemo_config.precision = precision
     print(f"nemo_config: {nemo_config}")
 
-    trainer = Trainer(plugins=plugins, accelerator='cpu', precision=precision, strategy=NLPDDPStrategy())
+    trainer = Trainer(plugins=plugins, accelerator='cpu', strategy=NLPDDPStrategy())
 
     hidden_size = nemo_config.hidden_size
     head_num = nemo_config.num_attention_heads

diff --git a/scripts/checkpoint_converters/convert_starcoder2_hf_to_nemo.py b/scripts/checkpoint_converters/convert_starcoder2_hf_to_nemo.py
@@ -194,7 +194,7 @@ def convert(args):
     nemo_config.precision = precision
     logging.info(f"nemo_config: {nemo_config}")
 
-    trainer = Trainer(plugins=plugins, accelerator='cpu', precision=precision, strategy=NLPDDPStrategy())
+    trainer = Trainer(plugins=plugins, accelerator='cpu', strategy=NLPDDPStrategy())
 
     hidden_size = nemo_config.hidden_size
     head_num = nemo_config.num_attention_heads