Chranos/enginex-mlu370-vllm
1
0
Fork 0
forked from EngineX-Cambricon/enginex-mlu370-vllm
Code Pull Requests Activity

add dynamic register

Browse Source
This commit is contained in:
Chranos
2026-02-05 15:53:43 +08:00
parent 9563c9af0d
commit 92f0016e6f
4 changed files with 244 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
vllm-v0.6.2/vllm/config.py
View File

@@ -274,7 +274,13 @@ class ModelConfig:
self, limit_mm_per_prompt: Optional[Mapping[str, int]]
) -> Optional["MultiModalConfig"]:
architectures = getattr(self.hf_config, "architectures", [])
if ModelRegistry.is_multimodal_model(architectures):
if ModelRegistry.is_multimodal_model(
architectures,
model_path=self.model,
revision=self.revision,
trust_remote_code=self.trust_remote_code,
hf_config=self.hf_config,
):
return MultiModalConfig(limit_per_prompt=limit_mm_per_prompt or {})
if limit_mm_per_prompt:
@@ -308,11 +314,23 @@ class ModelConfig:
def _init_attention_free(self) -> bool:
architectures = getattr(self.hf_config, "architectures", [])
return ModelRegistry.is_attention_free_model(architectures)
return ModelRegistry.is_attention_free_model(
architectures,
model_path=self.model,
revision=self.revision,
trust_remote_code=self.trust_remote_code,
hf_config=self.hf_config,
)
def _init_has_inner_state(self) -> bool:
architectures = getattr(self.hf_config, "architectures", [])
return ModelRegistry.model_has_inner_state(architectures)
return ModelRegistry.model_has_inner_state(
architectures,
model_path=self.model,
revision=self.revision,
trust_remote_code=self.trust_remote_code,
hf_config=self.hf_config,
)
def _verify_tokenizer_mode(self) -> None:
tokenizer_mode = self.tokenizer_mode.lower()