mmvq in cuda path

Author: airMeng

ggml/src/ggml-sycl.cpp

@@ -3581,7 +3581,8 @@ static void ggml_sycl_mul_mat(ggml_backend_sycl_context & ctx, const ggml_tensor
 
     bool use_mul_mat_vec_q =  ggml_is_quantized(src0->type)
         && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32
-        && src1->ne[1] <= MMVQ_MAX_BATCH_SIZE && src1->ne[1] > MMVQ_MIN_BATCH_SIZE;
+        && src1->ne[1] <= MMVQ_MAX_BATCH_SIZE
+        && ctx.stream()->get_backend() == sycl::backend::ext_oneapi_cuda ? true: src1->ne[1] > MMVQ_MIN_BATCH_SIZE;
 
     bool use_mul_mat_q =  ggml_sycl_supports_mmq(src0->type)
         && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32;