Restore the effect of liger_kernel's monkey_patch on global modules in UT. #3680
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…t of liger_kernel's monkey_patch on global modules.
|
@kashif and @shirinyamani pls help review. |
|
actually your approach is also good... perhaps we can just use it with the @parameterized.expand(list(itertools.product(MODELS_TO_TEST, PACKING_OPTIONS)))
@require_liger_kernel
def test_sft_trainer_with_liger(self, model_name, packing):
"""
Tests if passing use_liger=True to SFTConfig loads and runs the trainer with AutoLigerKernelForCausalLM as
expected.
"""
import importlib
def cleanup_liger_patches(trainer):
"""Clean up liger_kernel patches by reloading the model's specific module"""
try:
# Get the specific module that was used by the trainer's model
module_path = trainer.model.__module__
reload_module = importlib.import_module(module_path)
importlib.reload(reload_module)
except Exception:
pass # Continue if reload fails
with tempfile.TemporaryDirectory() as tmp_dir:
training_args = SFTConfig(
output_dir=tmp_dir,
logging_strategy="no",
report_to="none",
per_device_train_batch_size=2,
max_steps=2,
packing=packing,
max_length=self.max_length,
use_liger_kernel=True,
)
trainer = SFTTrainer(
model_name,
args=training_args,
train_dataset=self.train_dataset,
eval_dataset=self.eval_dataset,
)
# Register cleanup now that we have the trainer
self.addCleanup(cleanup_liger_patches, trainer)
trainer.train()
release_memory(trainer.model, trainer) |
Great idea, thanks for the suggestion! |
kashif
approved these changes
Jul 5, 2025
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
marcandrelarochelle
pushed a commit
to marcandrelarochelle/trl
that referenced
this pull request
Jul 29, 2025
…n UT. (huggingface#3680) Co-authored-by: Kashif Rasul <kashif.rasul@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
The liger_kernel modifies the global modules configuration. When performing UT in the same process, the impact of liger_kernel's monkey_patch should be promptly restored after testing.
Reproduce the issue
Based on the implementation of
tests/slow/test_sft_slow.py::test_sft_trainer_with_liger, I wrote a simple script to demonstrate the issue.The script output is as follows:
Testing Done
If these changes are not manually reverted, they will affect the loading of related models in subsequent operations within the same process. Reloading the modules involved in liger can restore the global configuration.