EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

misc: rm unused model_loader (#1110)

Browse Source
This commit is contained in:
Yineng Zhang
2024-08-15 23:29:35 +08:00
committed by GitHub
parent 14cb544d56
commit 9195d1362a
3 changed files with 1 additions and 576 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
python/sglang/srt/model_executor/model_runner.py
View File

@@ -38,6 +38,7 @@ from vllm.distributed import (
init_distributed_environment,
initialize_model_parallel,
)
from vllm.model_executor.model_loader import get_model
from vllm.model_executor.models import ModelRegistry
from sglang.global_config import global_config
@@ -168,15 +169,6 @@ class ModelRunner:
if self.model_config.model_overide_args is not None:
vllm_model_config.hf_config.update(self.model_config.model_overide_args)
if (
self.server_args.efficient_weight_load
and "llama" in self.server_args.model_path.lower()
and self.server_args.quantization == "fp8"
):
from sglang.srt.model_loader.model_loader import get_model
else:
from vllm.model_executor.model_loader import get_model
self.model = get_model(
model_config=vllm_model_config,
device_config=device_config,