SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Split checkpoint from model_name_or_path in examples (#11492)

Browse Source 
* Split checkpoint from model_name_or_path in examples

* Address review comments

* Address review comments
This commit is contained in:
Sylvain Gugger
2021-04-29 18:33:47 -04:00
committed by GitHub
parent d6ec54ba36
commit b29eb247d3
13 changed files with 62 additions and 69 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/pytorch/README.md
View File

@@ -65,7 +65,7 @@ examples/pytorch/token-classification/run_ner.py -h
You can resume training from a previous checkpoint like this:
1. Pass `--output_dir previous_output_dir` without `--overwrite_output_dir` to resume training from the latest checkpoint in `output_dir` (what you would use if the training was interrupted, for instance).
2. Pass `--model_name_or_path path_to_a_specific_checkpoint` to resume training from that checkpoint folder.
2. Pass `--resume_from_checkpoint path_to_a_specific_checkpoint` to resume training from that checkpoint folder.
Should you want to turn an example into a notebook where you'd no longer have access to the command
line, 🤗 Trainer supports resuming from a checkpoint via `trainer.train(resume_from_checkpoint)`.

11
examples/pytorch/language-modeling/run_clm.py
View File

@@ -190,7 +190,7 @@ def main():
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
@@ -413,12 +413,11 @@ def main():
# Training
if training_args.do_train:
if last_checkpoint is not None:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif model_args.model_name_or_path is not None and os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload

11
examples/pytorch/language-modeling/run_mlm.py
View File

@@ -199,7 +199,7 @@ def main():
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
@@ -443,12 +443,11 @@ def main():
# Training
if training_args.do_train:
if last_checkpoint is not None:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif model_args.model_name_or_path is not None and os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload
metrics = train_result.metrics

11
examples/pytorch/language-modeling/run_plm.py
View File

@@ -196,7 +196,7 @@ def main():
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
@@ -419,12 +419,11 @@ def main():
# Training
if training_args.do_train:
if last_checkpoint is not None:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif model_args.model_name_or_path is not None and os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload
metrics = train_result.metrics

11
examples/pytorch/multiple-choice/run_swag.py
View File

@@ -223,7 +223,7 @@ def main():
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
@@ -398,12 +398,11 @@ def main():
# Training
if training_args.do_train:
if last_checkpoint is not None:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload
metrics = train_result.metrics

11
examples/pytorch/question-answering/run_qa.py
View File

@@ -216,7 +216,7 @@ def main():
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
@@ -557,12 +557,11 @@ def main():
# Training
if training_args.do_train:
if last_checkpoint is not None:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload

11
examples/pytorch/question-answering/run_qa_beam_search.py
View File

@@ -215,7 +215,7 @@ def main():
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
@@ -595,12 +595,11 @@ def main():
# Training
if training_args.do_train:
if last_checkpoint is not None:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload

11
examples/pytorch/summarization/run_summarization.py
View File

@@ -272,7 +272,7 @@ def main():
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
@@ -520,12 +520,11 @@ def main():
# Training
if training_args.do_train:
if last_checkpoint is not None:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload

12
examples/pytorch/text-classification/run_glue.py
View File

@@ -196,7 +196,7 @@ def main():
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
@@ -448,14 +448,10 @@ def main():
# Training
if training_args.do_train:
checkpoint = None
if last_checkpoint is not None:
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
# Check the config from that potential checkpoint has the right number of labels before using it as a
# checkpoint.
if AutoConfig.from_pretrained(model_args.model_name_or_path).num_labels == num_labels:
checkpoint = model_args.model_name_or_path
train_result = trainer.train(resume_from_checkpoint=checkpoint)
metrics = train_result.metrics
max_train_samples = (

9
examples/pytorch/text-classification/run_xnli.py
View File

@@ -335,13 +335,10 @@ def main():
# Training
if training_args.do_train:
checkpoint = None
if last_checkpoint is not None:
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
# Check the config from that potential checkpoint has the right number of labels before using it as a
# checkpoint.
if AutoConfig.from_pretrained(model_args.model_name_or_path).num_labels == num_labels:
checkpoint = model_args.model_name_or_path
train_result = trainer.train(resume_from_checkpoint=checkpoint)
metrics = train_result.metrics
max_train_samples = (

11
examples/pytorch/token-classification/run_ner.py
View File

@@ -189,7 +189,7 @@ def main():
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
@@ -437,12 +437,11 @@ def main():
# Training
if training_args.do_train:
if last_checkpoint is not None:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
metrics = train_result.metrics
trainer.save_model() # Saves the tokenizer too for easy upload

11
examples/pytorch/translation/run_translation.py
View File

@@ -256,7 +256,7 @@ def main():
f"Output directory ({training_args.output_dir}) already exists and is not empty. "
"Use --overwrite_output_dir to overcome."
)
elif last_checkpoint is not None:
elif last_checkpoint is not None and training_args.resume_from_checkpoint is None:
logger.info(
f"Checkpoint detected, resuming training at {last_checkpoint}. To avoid this behavior, change "
"the `--output_dir` or add `--overwrite_output_dir` to train from scratch."
@@ -512,12 +512,11 @@ def main():
# Training
if training_args.do_train:
if last_checkpoint is not None:
checkpoint = None
if training_args.resume_from_checkpoint is not None:
checkpoint = training_args.resume_from_checkpoint
elif last_checkpoint is not None:
checkpoint = last_checkpoint
elif os.path.isdir(model_args.model_name_or_path):
checkpoint = model_args.model_name_or_path
else:
checkpoint = None
train_result = trainer.train(resume_from_checkpoint=checkpoint)
trainer.save_model() # Saves the tokenizer too for easy upload