init v0.11.0rc0

2025-10-14 10:38:28 +08:00
parent 67afd0ea78
commit 66dc16f966
278 changed files with 28130 additions and 11708 deletions
--- a/vllm_ascend/ops/activation.py
+++ b/vllm_ascend/ops/activation.py
@@ -35,8 +35,10 @@ class AscendSiluAndMul(SiluAndMul):

        from vllm_ascend.utils import is_310p

+        torch.ops.vllm.maybe_prefetch_mlp_down_proj(x)
        if is_310p():
            out = torch_npu.npu_swiglu(x.to(torch.float32)).to(torch.float16)
        else:
            out = torch_npu.npu_swiglu(x)
+        torch.ops.vllm.maybe_wait_prefetch_done(out)
        return out