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

Update all no_trainer with skip_first_batches (#23664)

Browse Source
This commit is contained in:
Zachary Mueller
2023-05-22 14:49:31 -04:00
committed by GitHub
parent 26a06814a1
commit b191d7db44
12 changed files with 115 additions and 98 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
examples/pytorch/image-classification/run_image_classification_no_trainer.py
View File

@@ -451,22 +451,26 @@ def main():
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
resume_step = int(training_difference.replace("step_", ""))
starting_epoch = resume_step // len(train_dataloader)
resume_step -= starting_epoch * len(train_dataloader)
completed_steps = resume_step
# update the progress_bar if load from checkpoint
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
with accelerator.accumulate(model):
outputs = model(**batch)
loss = outputs.loss

20
examples/pytorch/image-pretraining/run_mim_no_trainer.py
View File

@@ -660,29 +660,27 @@ def main():
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
# need to multiply `gradient_accumulation_steps` to reflect real steps
resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
starting_epoch = resume_step // len(train_dataloader)
resume_step -= starting_epoch * len(train_dataloader)
completed_steps = resume_step
# update the progress_bar if load from checkpoint
progress_bar.update(starting_epoch * num_update_steps_per_epoch)
completed_steps = starting_epoch * num_update_steps_per_epoch
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
if step % args.gradient_accumulation_steps == 0:
progress_bar.update(1)
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
with accelerator.accumulate(model):
outputs = model(**batch)
loss = outputs.loss

20
examples/pytorch/language-modeling/run_clm_no_trainer.py
View File

@@ -566,29 +566,27 @@ def main():
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
# need to multiply `gradient_accumulation_steps` to reflect real steps
resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
starting_epoch = resume_step // len(train_dataloader)
resume_step -= starting_epoch * len(train_dataloader)
completed_steps = resume_step
# update the progress_bar if load from checkpoint
progress_bar.update(starting_epoch * num_update_steps_per_epoch)
completed_steps = starting_epoch * num_update_steps_per_epoch
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
if step % args.gradient_accumulation_steps == 0:
progress_bar.update(1)
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
with accelerator.accumulate(model):
outputs = model(**batch)
loss = outputs.loss

20
examples/pytorch/language-modeling/run_mlm_no_trainer.py
View File

@@ -610,29 +610,27 @@ def main():
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
# need to multiply `gradient_accumulation_steps` to reflect real steps
resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
starting_epoch = resume_step // len(train_dataloader)
resume_step -= starting_epoch * len(train_dataloader)
completed_steps = resume_step
# update the progress_bar if load from checkpoint
progress_bar.update(starting_epoch * num_update_steps_per_epoch)
completed_steps = starting_epoch * num_update_steps_per_epoch
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
if step % args.gradient_accumulation_steps == 0:
progress_bar.update(1)
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
with accelerator.accumulate(model):
outputs = model(**batch)
loss = outputs.loss

18
examples/pytorch/multiple-choice/run_swag_no_trainer.py
View File

@@ -557,22 +557,26 @@ def main():
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
resume_step = int(training_difference.replace("step_", ""))
starting_epoch = resume_step // len(train_dataloader)
resume_step -= starting_epoch * len(train_dataloader)
completed_steps = resume_step
# update the progress_bar if load from checkpoint
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
with accelerator.accumulate(model):
outputs = model(**batch)
loss = outputs.loss

18
examples/pytorch/question-answering/run_qa_beam_search_no_trainer.py
View File

@@ -809,22 +809,26 @@ def main():
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
resume_step = int(training_difference.replace("step_", ""))
starting_epoch = resume_step // len(train_dataloader)
resume_step -= starting_epoch * len(train_dataloader)
completed_steps = resume_step
# update the progress_bar if load from checkpoint
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
with accelerator.accumulate(model):
outputs = model(**batch)
loss = outputs.loss

18
examples/pytorch/question-answering/run_qa_no_trainer.py
View File

@@ -825,22 +825,26 @@ def main():
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
resume_step = int(training_difference.replace("step_", ""))
starting_epoch = resume_step // len(train_dataloader)
resume_step -= starting_epoch * len(train_dataloader)
completed_steps = resume_step
# update the progress_bar if load from checkpoint
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
with accelerator.accumulate(model):
outputs = model(**batch)
loss = outputs.loss

20
examples/pytorch/semantic-segmentation/run_semantic_segmentation_no_trainer.py
View File

@@ -554,22 +554,26 @@ def main():
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
resume_step = int(training_difference.replace("step_", ""))
starting_epoch = resume_step // len(train_dataloader)
resume_step -= starting_epoch * len(train_dataloader)
completed_steps = resume_step
# update the progress_bar if load from checkpoint
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
model.train()
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
with accelerator.accumulate(model):
outputs = model(**batch)
loss = outputs.loss

18
examples/pytorch/summarization/run_summarization_no_trainer.py
View File

@@ -626,22 +626,26 @@ def main():
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
resume_step = int(training_difference.replace("step_", ""))
starting_epoch = resume_step // len(train_dataloader)
resume_step -= starting_epoch * len(train_dataloader)
completed_steps = resume_step
# update the progress_bar if load from checkpoint
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
with accelerator.accumulate(model):
outputs = model(**batch)
loss = outputs.loss

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

@@ -510,12 +510,12 @@ def main():
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
outputs = model(**batch)
loss = outputs.loss
# We keep track of the loss at each epoch

12
examples/pytorch/token-classification/run_ner_no_trainer.py
View File

@@ -668,12 +668,12 @@ def main():
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
outputs = model(**batch)
loss = outputs.loss
# We keep track of the loss at each epoch

19
examples/pytorch/translation/run_translation_no_trainer.py
View File

@@ -607,28 +607,27 @@ def main():
if "epoch" in training_difference:
starting_epoch = int(training_difference.replace("epoch_", "")) + 1
resume_step = None
completed_steps = starting_epoch * num_update_steps_per_epoch
else:
# need to multiply `gradient_accumulation_steps` to reflect real steps
resume_step = int(training_difference.replace("step_", "")) * args.gradient_accumulation_steps
starting_epoch = resume_step // len(train_dataloader)
resume_step -= starting_epoch * len(train_dataloader)
completed_steps = resume_step
# update the progress_bar if load from checkpoint
progress_bar.update(starting_epoch * num_update_steps_per_epoch)
completed_steps = starting_epoch * num_update_steps_per_epoch
progress_bar.update(completed_steps)
for epoch in range(starting_epoch, args.num_train_epochs):
model.train()
if args.with_tracking:
total_loss = 0
for step, batch in enumerate(train_dataloader):
# We need to skip steps until we reach the resumed step
if args.resume_from_checkpoint and epoch == starting_epoch:
if resume_step is not None and step < resume_step:
if step % args.gradient_accumulation_steps == 0:
progress_bar.update(1)
completed_steps += 1
continue
if args.resume_from_checkpoint and epoch == starting_epoch and resume_step is not None:
# We skip the first `n` batches in the dataloader when resuming from a checkpoint
active_dataloader = accelerator.skip_first_batches(train_dataloader, resume_step)
else:
active_dataloader = train_dataloader
for step, batch in enumerate(active_dataloader):
outputs = model(**batch)
loss = outputs.loss
# We keep track of the loss at each epoch