Upgrade to vllm 0.17.0 corex v4.1 overlay

2026-04-29 19:38:22 +08:00
parent 8fac6062e4
commit 938d0854a5
430 changed files with 35969 additions and 14511 deletions
--- a/vllm/model_executor/layers/quantization/gptq.py
+++ b/vllm/model_executor/layers/quantization/gptq.py
@@ -128,7 +128,7 @@ class GPTQConfig(QuantizationConfig):

    @classmethod
    def get_supported_act_dtypes(cls) -> list[torch.dtype]:
-        return [torch.half, torch.bfloat16]
+        return [torch.bfloat16, torch.half]

    @classmethod
    # Need to figure it out