diff --git a/ggml/src/ggml-cuda/mmq.cuh b/ggml/src/ggml-cuda/mmq.cuh index 209555ce39..5424ba79b5 100644 --- a/ggml/src/ggml-cuda/mmq.cuh +++ b/ggml/src/ggml-cuda/mmq.cuh @@ -401,7 +401,6 @@ static __device__ __forceinline__ void vec_dot_q4_0_q8_1_dp4a( sum[j0/nwarps*mmq_y/warp_size + i0/warp_size] += vec_dot_q4_0_q8_1_impl (&x_qs[i*(MMQ_TILE_NE_K + 1) + k0/QR4_0], u, x_df[i*(MMQ_TILE_NE_K/QI4_0) + i/QI4_0 + k0/(QR4_0*QI4_0)], y_ds[j*MMQ_TILE_Y_K + k01/QI8_1]); - } } }