EngineX-Ascend/enginex-ascend-910-vllm
10
0
Fork 1
Code Issues Pull Requests Actions Projects Releases Wiki Activity

init v0.11.0rc0

Browse Source
This commit is contained in:
luopingyi
2025-10-14 10:38:28 +08:00
parent 67afd0ea78
commit 66dc16f966
278 changed files with 28130 additions and 11708 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
vllm_ascend/models/qwen2_vl.py
View File

@@ -40,6 +40,8 @@ from vllm.model_executor.models.qwen2_vl import (
from vllm.model_executor.models.utils import maybe_prefix
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm_ascend.utils import vllm_version_is
MIN_PAD_SIZE = 64 # min_size to pad weight
MAX_PAD_SIZE = 128 # max_size to pad weight
@@ -343,10 +345,18 @@ class AscendQwen2VLForConditionalGeneration(Qwen2VLForConditionalGeneration):
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
super().__init__(vllm_config=vllm_config, prefix=prefix)
self.visual = AscendQwen2VisionTransformer(
self.config.vision_config,
norm_eps=getattr(self.config, "rms_norm_eps", 1e-6),
quant_config=self._maybe_ignore_quant_config(
vllm_config.quant_config),
prefix=maybe_prefix(prefix, "visual"),
)
if vllm_version_is("0.10.2"):
self.visual = AscendQwen2VisionTransformer(
self.config.vision_config,
norm_eps=getattr(self.config, "rms_norm_eps", 1e-6),
quant_config=self._maybe_ignore_quant_config(
vllm_config.quant_config),
prefix=maybe_prefix(prefix, "visual"),
)
else:
self.visual = AscendQwen2VisionTransformer(
self.config.vision_config,
norm_eps=getattr(self.config, "rms_norm_eps", 1e-6),
quant_config=vllm_config.quant_config,
prefix=maybe_prefix(prefix, "visual"),
)