[Bugfix] Retry config read to survive concurrent HF cache refresh (#49001)
Signed-off-by: pei.zhang <pei.zhang@amd.com> Co-authored-by: Claude <noreply@anthropic.com>
peizhang56 committed
Jul 22, 2026 at 15:35 UTC
7d10a4cfce45527e03e9d6ef0f1a8c286256b008
1 file changed
+11
-7
vllm/transformers_utils/config.py
+11
-7
@@ -731,13 +731,17 @@ def get_config(
731
raise ValueError(error_message) from e
732
733
config_parser = get_config_parser(config_format)
734
- config_dict, config = config_parser.parse(
735
- model,
736
- trust_remote_code=trust_remote_code,
737
- revision=revision,
738
- code_revision=code_revision,
739
- hf_overrides=hf_overrides_kw or hf_overrides_fn,
740
- **kwargs,
734
+ # Retry to tolerate a concurrent HF cache refresh briefly hiding config.json.
735
+ config_dict, config = with_retry(
736
+ lambda: config_parser.parse(
737
+ model,
738
+ trust_remote_code=trust_remote_code,
739
+ revision=revision,
740
+ code_revision=code_revision,
741
+ hf_overrides=hf_overrides_kw or hf_overrides_fn,
742
+ **kwargs,
743
+ ),
744
+ f"Error parsing config for {model}",
745
)
746
747
# Architecture mapping for models without explicit architectures field