xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-03-28 10:39:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

update with main

Browse Source 
Signed-off-by: Xingyu Liu <charlotteliu12x@gmail.com>
This commit is contained in:
Xingyu Liu 2025-12-16 11:58:06 -08:00
parent 9a75278f2a
commit 0f85cfdc65
1 changed files with 13 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
vllm/transformers_utils/model_arch_config_convertor.py
View File

@ -26,6 +26,9 @@ class ModelArchConfigConvertorBase:
self.hf_config = hf_config
self.hf_text_config = hf_text_config
def get_architectures(self) -> list[str]:
return getattr(self.hf_config, "architectures", [])
def get_num_hidden_layers(self) -> int:
return getattr(self.hf_text_config, "num_hidden_layers", 0)
@ -240,7 +243,7 @@ class ModelArchConfigConvertorBase:
def convert(self, model_id: str, revision: str | None) -> ModelArchitectureConfig:
model_arch_config = ModelArchitectureConfig(
architectures=getattr(self.hf_config, "architectures", []),
architectures=self.get_architectures(),
model_type=self.hf_config.model_type,
text_model_type=getattr(self.hf_text_config, "model_type", None),
hidden_size=self.get_hidden_size(),
@ -331,7 +334,15 @@ class NemotronNasModelArchConfigConvertor(ModelArchConfigConvertorBase):
self.hf_text_config.num_attention_heads
// block.attention.n_heads_in_group
)
raise RuntimeError("Couldn't determine number of kv heads")
raise RuntimeError(
"Could not determine the number of key-value attention heads "
"from model configuration. "
f"Architecture: {self.get_architectures()}. "
"This usually indicates an unsupported model architecture or "
"missing configuration. "
"Please check if your model is supported at: "
"https://docs.vllm.ai/en/latest/models/supported_models.html"
)
class DeepSeekMTPModelArchConfigConvertor(ModelArchConfigConvertorBase):