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

[FlashInfer] Upgrade to 0.2.0 (#11194)

Browse Source 
Signed-off-by: Bowen Wang <abmfy@icloud.com>
Signed-off-by: youkaichao <youkaichao@gmail.com>
Co-authored-by: youkaichao <youkaichao@gmail.com>
This commit is contained in:
Bowen Wang
2025-01-28 02:19:24 +08:00
committed by GitHub
parent 3f1fc7425a
commit 2bc3fbba0c
10 changed files with 256 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/model_executor/model_loader/loader.py
View File

@@ -114,7 +114,7 @@ def _initialize_model(
all_params = [param.name for param in signatures.parameters.values()]
if "vllm_config" in all_params and "prefix" in all_params:
# new-style model class
with set_current_vllm_config(vllm_config):
with set_current_vllm_config(vllm_config, check_compile=True):
return model_class(vllm_config=vllm_config, prefix=prefix)
msg = ("vLLM model class should accept `vllm_config` and `prefix` as "
@@ -142,7 +142,7 @@ def _initialize_model(
kwargs["lora_config"] = vllm_config.lora_config
if "scheduler_config" in all_params:
kwargs["scheduler_config"] = vllm_config.scheduler_config
with set_current_vllm_config(vllm_config):
with set_current_vllm_config(vllm_config, check_compile=True):
return model_class(**kwargs)

3
vllm/model_executor/model_loader/tensorizer.py
View File

@@ -288,7 +288,8 @@ class TensorizerAgent:
model_args.torch_dtype = self.tensorizer_config.dtype
assert self.tensorizer_config.model_class is not None
# TODO: Do we need to consider old-style model class?
with no_init_or_tensor(), set_current_vllm_config(self.vllm_config):
with no_init_or_tensor(), set_current_vllm_config(self.vllm_config,
check_compile=True):
return self.tensorizer_config.model_class(
vllm_config=self.vllm_config, )