remove row=1 cond (llama/6532)

diff --git a/src/ggml-sycl.cpp b/src/ggml-sycl.cpp
index b83881496e4d4621f79b627aa9955a00164e9902..55a1eedb553b1a8ba667d2b6774bc5ebc3650e7d 100644 (file)
--- a/src/ggml-sycl.cpp
+++ b/src/ggml-sycl.cpp
@@ -15776,7 +15776,7 @@ static void ggml_sycl_mul_mat(const ggml_tensor * src0, const ggml_tensor * src1
 #ifdef GGML_SYCL_FORCE_DMMV
             const bool use_mul_mat_vec_q = false;
 #else
-            bool use_mul_mat_vec_q = min_compute_capability >= VER_4VEC && ggml_is_quantized(src0->type) && ggml_nrows(src1) == 1;
+            bool use_mul_mat_vec_q = min_compute_capability >= VER_4VEC && ggml_is_quantized(src0->type);
             use_mul_mat_vec_q = use_mul_mat_vec_q ||
                 (src0->type == GGML_TYPE_IQ2_XXS) || (src0->type == GGML_TYPE_IQ2_XS) || (src0->type == GGML_TYPE_IQ2_S) ||
                 (src0->type == GGML_TYPE_IQ3_XXS) || (src0->type == GGML_TYPE_IQ3_S) ||
@@ -15787,7 +15787,6 @@ static void ggml_sycl_mul_mat(const ggml_tensor * src0, const ggml_tensor * src1
 #endif // GGML_SYCL_FORCE_DMMV
 
             if (use_mul_mat_vec_q) {
-                // NOTE: this kernel does not support ggml_nrows(src1) > 1
                 // GGML_SYCL_DEBUG("ggml_sycl_mul_mat ggml_sycl_op_mul_mat_vec_q path\n");
                 ggml_sycl_op_mul_mat(src0, src1, dst, ggml_sycl_op_mul_mat_vec_q, true);
             } else {