biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Quantization] Add field to skip unquantized modules for GPTQ config (#25455)

Browse Source 
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
This commit is contained in:
Isotr0py
2025-09-26 23:47:41 +08:00
committed by GitHub
parent db1e42f627
commit d4d9899860
16 changed files with 219 additions and 153 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
vllm/model_executor/models/qwen3_vl.py
View File

@@ -50,9 +50,6 @@ 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.gptq import GPTQConfig
from vllm.model_executor.layers.quantization.gptq_marlin import (
GPTQMarlinConfig)
from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.model_executor.models.module_mapping import MultiModelKeys
@@ -1058,7 +1055,7 @@ class Qwen3VLForConditionalGeneration(nn.Module, SupportsMultiModal,
self.visual = Qwen3_VisionTransformer(
config.vision_config,
norm_eps=getattr(config, "rms_norm_eps", 1e-6),
quant_config=self._maybe_ignore_quant_config(quant_config),
quant_config=quant_config,
prefix=maybe_prefix(prefix, "visual"),
use_data_parallel=self.use_data_parallel,
)
@@ -1116,13 +1113,6 @@ class Qwen3VLForConditionalGeneration(nn.Module, SupportsMultiModal,
for idx in range(self.deepstack_num_level):
self.deepstack_input_embeds[idx][:num_tokens].zero_()
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.
if isinstance(quant_config, (GPTQConfig, GPTQMarlinConfig)):
return None
return quant_config
def _validate_and_reshape_mm_tensor(self, mm_input: object,
name: str) -> torch.Tensor:
if not isinstance(mm_input, (torch.Tensor, list)):