chore(qwen2): display warning log only when sliding window attention …

[edwardzjl]

…is enabled

What does this PR do?

In #36162, a tiny refactor has been made to qwen2:

diff --git a/src/transformers/models/qwen2/configuration_qwen2.py b/src/transformers/models/qwen2/configuration_qwen2.py
index 16979865e4..2e82f1976f 100644
--- a/src/transformers/models/qwen2/configuration_qwen2.py
+++ b/src/transformers/models/qwen2/configuration_qwen2.py
@@ -174,7 +174,7 @@ class Qwen2Config(PretrainedConfig):
         self.num_hidden_layers = num_hidden_layers
         self.num_attention_heads = num_attention_heads
         self.use_sliding_window = use_sliding_window
-        self.sliding_window = sliding_window if use_sliding_window else None
+        self.sliding_window = sliding_window  # we check `use_sliding_window` in the modeling code
         self.max_window_layers = max_window_layers

         # for backward compatibility

However, during modeling, this sliding_window attribute will trigger a warning log, even if it is not enabled:

transformers/src/transformers/models/qwen2/modeling_qwen2.py

Lines 237 to 241 in e18f233

	if config.sliding_window and config._attn_implementation != "flash_attention_2":
	logger.warning_once(
	f"Sliding Window Attention is enabled but not implemented for `{config._attn_implementation}`; "
	"unexpected results may be encountered."
	)

This PR could prevent this warning message being displayed if sliding window is not enabled, which reduces unnecessary confusion.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

@ArthurZucker

[edwardzjl]

@ArthurZucker Sorry to bother, but I'm not sure if I am doing it correctly.

According to the header:

transformers/src/transformers/models/qwen2/modeling_qwen2.py

Lines 1 to 6 in e18f233

	# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
	# This file was automatically generated from src/transformers/models/qwen2/modular_qwen2.py.
	# Do NOT edit this file manually as any edits will be overwritten by the generation of
	# the file from the modular. If any change should be done, please apply the change to the
	# modular_qwen2.py file directly. One of our CI enforces this.
	# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨

I should not edit the modeling_qwen2.py directly, so I made the changes to the modular_qwen2.py. However, it seems that the CI is failing because the modeling_qwen2.py is not consistent with modular_qwen2.py:

https://app.circleci.com/pipelines/github/huggingface/transformers/119338/workflows/c1ea9aee-f73c-463d-b1c2-514cb89d4b72/jobs/1612428?utm_campaign=vcs-integration-link&utm_medium=referral&utm_source=github-checks-link&utm_content=summary