Skip to content
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

Jump to bottom

update to be deprecated evaluation_strategy #1682

Merged
merged 3 commits into from
Nov 16, 2024
Merged

update to be deprecated evaluation_strategy #1682

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
8fda363
update to be deprecated evaluation_strategy and c4 dataset
winglian Jun 3, 2024
fc09867
chore: lint
winglian Nov 14, 2024
7da216d
remap eval strategy to new config and add tests
winglian Nov 15, 2024
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
16 changes: 7 additions & 9 deletions src/axolotl/core/trainer_builder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1416,17 +1416,15 @@ def build(self, total_num_steps):

if not self.cfg.test_datasets and self.cfg.val_set_size == 0:
# no eval set, so don't eval
training_arguments_kwargs["evaluation_strategy"] = "no"
training_arguments_kwargs["eval_strategy"] = "no"
elif self.cfg.eval_steps:
training_arguments_kwargs["evaluation_strategy"] = "steps"
training_arguments_kwargs["eval_strategy"] = "steps"
training_arguments_kwargs["eval_steps"] = self.cfg.eval_steps
elif self.cfg.evaluation_strategy:
training_arguments_kwargs[
"evaluation_strategy"
] = self.cfg.evaluation_strategy
elif self.cfg.eval_strategy:
training_arguments_kwargs["eval_strategy"] = self.cfg.eval_strategy
else:
# we have an eval set, but no steps defined, default to use epoch
training_arguments_kwargs["evaluation_strategy"] = "epoch"
training_arguments_kwargs["eval_strategy"] = "epoch"

if self.cfg.save_steps:
training_arguments_kwargs["save_strategy"] = "steps"
Expand Down Expand Up @@ -1860,10 +1858,10 @@ def build_training_arguments(self, total_num_steps):
training_args_kwargs["save_safetensors"] = self.cfg.save_safetensors

if self.eval_dataset:
training_args_kwargs["evaluation_strategy"] = "steps"
training_args_kwargs["eval_strategy"] = "steps"
training_args_kwargs["eval_steps"] = self.cfg.eval_steps
else:
training_args_kwargs["evaluation_strategy"] = "no"
training_args_kwargs["eval_strategy"] = "no"

if self.cfg.bf16 or self.cfg.bfloat16:
training_args_kwargs["bf16"] = True
Expand Down
5 changes: 1 addition & 4 deletions src/axolotl/utils/callbacks/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,10 +64,7 @@ def on_step_end(
control: TrainerControl,
**kwargs,
):
if (
args.evaluation_strategy == IntervalStrategy.STEPS
and state.global_step == 1
):
if args.eval_strategy == IntervalStrategy.STEPS and state.global_step == 1:
control.should_evaluate = True
return control

Expand Down
Loading
Loading