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

[Bugfix] Ignore GPTQ quantization of Qwen2-VL visual module #10169

Merged
merged 1 commit into from
Nov 9, 2024
Merged

[Bugfix] Ignore GPTQ quantization of Qwen2-VL visual module #10169

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
14 changes: 12 additions & 2 deletions vllm/model_executor/models/qwen2_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,9 @@
from vllm.model_executor.layers.linear import (ColumnParallelLinear,
RowParallelLinear)
from vllm.model_executor.layers.logits_processor import LogitsProcessor
from vllm.model_executor.layers.quantization import QuantizationConfig
from vllm.model_executor.layers.quantization import (GPTQConfig,
GPTQMarlinConfig,
QuantizationConfig)
from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
Expand Down Expand Up @@ -982,7 +984,7 @@ def __init__(self,
self.visual = Qwen2VisionTransformer(
config.vision_config,
norm_eps=getattr(config, "rms_norm_eps", 1e-6),
quant_config=quant_config,
quant_config=self._maybe_ignore_quant_config(quant_config),
Copy link
Collaborator

Choose a reason for hiding this comment

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

I have a question: What should we do if our users execute their own quantization, and they have quantized the visual encoder?

prefix="visual",
)

Expand All @@ -1008,6 +1010,14 @@ def __init__(self,
make_empty_intermediate_tensors_factory(
["hidden_states", "residual"], config.hidden_size))

def _maybe_ignore_quant_config(self, quant_config: QuantizationConfig):
# GPTQ configs do not have a list of ignored modules, however AutoGPTQ
# seems to avoid vision encoder sections for some models.
# See: https://huggingface.co/Qwen/Qwen2-VL-2B-Instruct-GPTQ-Int4
if isinstance(quant_config, (GPTQConfig, GPTQMarlinConfig)):
return None
return quant_config

def _validate_and_reshape_mm_tensor(self,
mm_input: Union[torch.Tensor,
List[torch.Tensor]],
Expand Down