diff --git a/python/sglang/srt/layers/attention/trtllm_mla_backend.py b/python/sglang/srt/layers/attention/trtllm_mla_backend.py index 4f81a46f0..bebe41037 100755 --- a/python/sglang/srt/layers/attention/trtllm_mla_backend.py +++ b/python/sglang/srt/layers/attention/trtllm_mla_backend.py @@ -1198,7 +1198,7 @@ class TRTLLMMLABackend(FlashInferMLAAttnBackend): out=out, ) else: - out = torch.empty( + out = torch.zeros( q.shape[0], q.shape[1], v.shape[2],