@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / ed70c70ea3
Fork 0

A high-throughput and memory-efficient inference and serving engine for LLMs

Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git

Synced Nov 17, 05:30 UTC

modelscope: fix issue when model parameter is not a model id but path of the model. (#2489)

liuyhwangyh committed Feb 7, 2024 at 01:57 UTC ed70c70ea3569670499717f06d117ed25ec32af4
1 file changed +6 -3
vllm/config.py
+6 -3
@@ -93,9 +93,12 @@ class ModelConfig:
93 # download model from ModelScope hub,
94 # lazy import so that modelscope is not required for normal use.
95 from modelscope.hub.snapshot_download import snapshot_download # pylint: disable=C
96 - model_path = snapshot_download(model_id=model,
97 - cache_dir=download_dir,
98 - revision=revision)
96 + if not os.path.exists(model):
97 + model_path = snapshot_download(model_id=model,
98 + cache_dir=download_dir,
99 + revision=revision)
100 + else:
101 + model_path = model
102 self.model = model_path
103 self.download_dir = model_path
104 self.tokenizer = model_path