diff --git a/vllm/models/kimi_k3/amd/ops/third_party/kda/chunk_intra.py b/vllm/models/kimi_k3/amd/ops/third_party/kda/chunk_intra.py index 0c600f3ba3..bc13318d83 100644 --- a/vllm/models/kimi_k3/amd/ops/third_party/kda/chunk_intra.py +++ b/vllm/models/kimi_k3/amd/ops/third_party/kda/chunk_intra.py @@ -349,7 +349,7 @@ def chunk_kda_fwd_kernel_inter_solve_fused( for num_warps in [1, 2, 4, 8] for num_stages in [2, 3, 4] ], - key=["BK", "NC", "BT", "HV"], + key=["BK", "BT", "HV"], ) @triton.jit(do_not_specialize=["B", "T"]) def chunk_kda_fwd_kernel_intra_sub_chunk( diff --git a/vllm/models/kimi_k3/nvidia/ops/third_party/kda/chunk_intra.py b/vllm/models/kimi_k3/nvidia/ops/third_party/kda/chunk_intra.py index 8b9aa86f9e..07b4a38893 100644 --- a/vllm/models/kimi_k3/nvidia/ops/third_party/kda/chunk_intra.py +++ b/vllm/models/kimi_k3/nvidia/ops/third_party/kda/chunk_intra.py @@ -451,7 +451,7 @@ def chunk_kda_fwd_kernel_inter_solve_fused( for num_warps in [1, 2, 4, 8] for num_stages in [2, 3, 4] ], - key=["BK", "NC", "BT", "HV"], + key=["BK", "BT", "HV"], ) @triton.jit(do_not_specialize=["B", "T"]) def chunk_kda_fwd_kernel_intra_sub_chunk(