Skip to content

fix dpo graph compile error in evaluation #630

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
regisss merged 1 commit into from
Jan 9, 2024
Merged

fix dpo graph compile error in evaluation #630

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 1 addition & 2 deletions optimum/habana/trl/trainer/dpo_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,7 +35,6 @@
DPODataCollatorWithPadding,
disable_dropout_in_model,
pad_to_length,
peft_module_casting_to_bf16,
)

from optimum.habana import GaudiConfig, GaudiTrainer, GaudiTrainingArguments
Expand Down Expand Up @@ -155,7 +154,7 @@ def make_inputs_require_grad(module, input, output):
# get peft model with the given config
model = get_peft_model(model, peft_config)
if args.bf16:
peft_module_casting_to_bf16(model)
model = model.to(torch.bfloat16)

# For models that use gradient_checkpoiting, we need to attach a hook that enables input
# to explicitly have `requires_grad=True`, otherwise training will either silently
Expand Down
3 changes: 1 addition & 2 deletions optimum/habana/trl/trainer/sft_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,6 @@
from trl.import_utils import is_peft_available
from trl.trainer.utils import (
DataCollatorForCompletionOnlyLM,
peft_module_casting_to_bf16,
)


Expand Down Expand Up @@ -143,7 +142,7 @@ def make_inputs_require_grad(module, input, output):

model = get_peft_model(model, peft_config)
if args.bf16:
peft_module_casting_to_bf16(model)
model = model.to(torch.bfloat16)

if tokenizer is None:
tokenizer = AutoTokenizer.from_pretrained(model.config._name_or_path)
Expand Down