diff --git a/.github/workflows/pr-test-npu.yml b/.github/workflows/pr-test-npu.yml index 9631a973a..57f04db6c 100644 --- a/.github/workflows/pr-test-npu.yml +++ b/.github/workflows/pr-test-npu.yml @@ -64,6 +64,7 @@ jobs: - ".github/workflows/pr-test-npu.yml" multimodal_gen: - "python/sglang/multimodal_gen/**/!(*.md|*.ipynb)" + - "python/sglang/jit_kernel/diffusion/triton/npu_fallback.py" - "python/sglang/srt/**" - "python/pyproject_npu.toml" - "scripts/ci/npu/npu_ci_install_dependency.sh" diff --git a/python/sglang/jit_kernel/diffusion/triton/npu_fallback.py b/python/sglang/jit_kernel/diffusion/triton/npu_fallback.py index 534ed47ea..f01ea841d 100644 --- a/python/sglang/jit_kernel/diffusion/triton/npu_fallback.py +++ b/python/sglang/jit_kernel/diffusion/triton/npu_fallback.py @@ -28,6 +28,7 @@ def apply_rotary_embedding_native( and x.dim() == 3 and x.shape[1] < NPU_ROTARY_MUL_MAX_NUM_HEADS and x.shape[2] < NPU_ROTARY_MUL_MAX_HEAD_SIZE + and not interleaved ): if cos.size(-1) * 2 == x.size(-1): cos = torch.cat([cos, cos], dim=-1)