diff --git a/ggml-cuda.cu b/ggml-cuda.cu index ef2b3d794..58b58f331 100644 --- a/ggml-cuda.cu +++ b/ggml-cuda.cu @@ -7387,7 +7387,6 @@ static void ggml_cuda_mul_mat_mat_batched_cublas(const ggml_tensor * src0, const dst->nb[2], dst->nb[3], r2, r3); CUDA_CHECK(cudaGetLastError()); - CUBLAS_CHECK(cublasSetStream(g_cublas_handles[id], main_stream)); CUBLAS_CHECK( cublasGemmBatchedEx(g_cublas_handles[id], CUBLAS_OP_T, CUBLAS_OP_N, ne01, ne11, ne10,