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/lora/ops/triton_ops/utils.py
+++ b/vllm/lora/ops/triton_ops/utils.py
@@ -316,3 +316,9 @@ def supports_pdl(device: torch.device | None = None) -> bool:
        and current_platform.has_device_capability(90)
        and not envs.VLLM_LORA_DISABLE_PDL
    )
+
+
+@lru_cache
+def supports_tma(device: torch.device | None = None) -> bool:
+    # TMA requires compute capability SM90 or above
+    return current_platform.is_cuda() and current_platform.has_device_capability(90)