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

[Bugfix] Fix tensor parallel for qwen2 classification model #10297

Merged
merged 3 commits into from
Nov 14, 2024
Merged

[Bugfix] Fix tensor parallel for qwen2 classification model #10297

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
a2509ed
fix qwen2cls tp
Isotr0py Nov 13, 2024
f443280
address comment
Isotr0py Nov 13, 2024
5cb22f5
score bias=False
Isotr0py Nov 13, 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
6 changes: 3 additions & 3 deletions tests/models/embedding/language/test_cls_models.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,14 +21,14 @@ def test_classification_models(
model: str,
dtype: str,
) -> None:
with vllm_runner(model, dtype=dtype) as vllm_model:
vllm_outputs = vllm_model.classify(example_prompts)

with hf_runner(model,
dtype=dtype,
auto_cls=AutoModelForSequenceClassification) as hf_model:
hf_outputs = hf_model.classify(example_prompts)

with vllm_runner(model, dtype=dtype) as vllm_model:
vllm_outputs = vllm_model.classify(example_prompts)

print(hf_outputs, vllm_outputs)

# check logits difference
Expand Down
7 changes: 6 additions & 1 deletion vllm/model_executor/models/qwen2_cls.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,9 +69,14 @@ def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
self.model = Qwen2Model(vllm_config=vllm_config,
prefix=maybe_prefix(prefix, "model"))

# hidden_states from Qwen2Model has been reduced,
# the input of score layer is not parallelized.
self.score = RowParallelLinear(config.hidden_size,
config.num_labels,
quant_config=quant_config)
quant_config=quant_config,
input_is_parallel=False,
bias=False,
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I just found that qwen2_cls doesn't have bias in score layer. 😅

The test should pass with this change. (The debugging was terrible 😅)

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah it passes now. Let's fix the timeout in another PR.

prefix=maybe_prefix(prefix, "score"))
self._pooler = Pooler.from_config_with_defaults(
pooler_config,
pooling_type=PoolingType.LAST,
Expand Down