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 16, 00:45 UTC
[Bugfix] Gracefully handle huggingface hub http error (#12571)
Roger Wang committed
Jan 31, 2025 at 00:19 UTC
7a8987dac5f0ed0c798a73e8b4ec8f5e640bc63a
1 file changed
+9
-1
vllm/transformers_utils/config.py
+9
-1
@@ -7,7 +7,8 @@ from typing import Any, Dict, Optional, Type, Union
7
import huggingface_hub
8
from huggingface_hub import (file_exists, hf_hub_download,
9
try_to_load_from_cache)
10
-from huggingface_hub.utils import (EntryNotFoundError, LocalEntryNotFoundError,
10
+from huggingface_hub.utils import (EntryNotFoundError, HfHubHTTPError,
11
+ LocalEntryNotFoundError,
12
RepositoryNotFoundError,
13
RevisionNotFoundError)
14
from torch import nn
@@ -294,6 +295,13 @@ def get_hf_file_to_dict(file_name: str,
295
logger.debug("File or repository not found in hf_hub_download",
296
e)
297
return None
298
+ except HfHubHTTPError as e:
299
+ logger.warning(
300
+ "Cannot connect to Hugging Face Hub. Skipping file "
301
+ "download for '%s':",
302
+ file_name,
303
+ exc_info=e)
304
+ return None
305
file_path = Path(hf_hub_file)
306
307
with open(file_path) as file: