sycl: fix mul_mat selection (#15092)

* webui: fix markdown table

* webui: fix table display with themes

ggml/src/ggml-sycl/ggml-sycl.cpp

@@ -2609,6 +2609,8 @@ static void ggml_sycl_mul_mat_vec_nc(ggml_backend_sycl_context & ctx, const ggml
     GGML_ASSERT(!ggml_backend_buffer_is_sycl_split(src0->buffer));
     GGML_ASSERT(src0->type == GGML_TYPE_F16);
     GGML_ASSERT(src1->type == GGML_TYPE_F32);
+    GGML_ASSERT(src1->ne[1] == 1);
+    GGML_ASSERT(src1->ne[3] == 1);
 
     const int64_t ne00 = src0->ne[0];
     const int64_t ne01 = src0->ne[1];
@@ -3196,7 +3198,7 @@ static void ggml_sycl_mul_mat(ggml_backend_sycl_context & ctx, const ggml_tensor
             // The kernel from the if path is faster for that specific case, but does not support all mul mats.
             ggml_sycl_mul_mat_batched_sycl(ctx, src0, src1, dst);
         }
-    } else if (!split && src0->type == GGML_TYPE_F16 && !ggml_is_contiguous(src0) && !ggml_is_transposed(src1) && src1->ne[1] == 1) {
+    } else if (!split && src0->type == GGML_TYPE_F16 && !ggml_is_contiguous(src0) && !ggml_is_transposed(src1) && src1->ne[1] == 1 && src1->ne[3] == 1) {
         // KQV single-batch
         ggml_sycl_mul_mat_vec_nc(ctx, src0, src1, dst);
     } else if (!split && src0->type == GGML_TYPE_F16 && !ggml_is_transposed(src0) && !ggml_is_transposed(src1) && src1->ne[2] * src1->ne[3] > 1) {

src/llama-model.cpp

@@ -13800,10 +13800,6 @@ struct llm_build_glm4_moe : public llm_graph_context {
                         LLM_FFN_SILU, LLM_FFN_PAR, il);
                 cb(cur, "ffn_out", il);
             } else {
-                // MoE layer with shared experts
-                const int64_t n_expert      = hparams.n_expert;
-                const int64_t n_expert_used = hparams.n_expert_used;
-
                 // Process routed experts using existing MoE infrastructure
                 ggml_tensor * routed_out = build_moe_ffn(cur,
                         model.layers[il].ffn_gate_inp,

tools/server/webui/src/index.scss

@@ -31,7 +31,24 @@ html {
   hr {
     @apply my-4 border-base-content/20 border-1;
   }
-  /* TODO: fix markdown table */
+  table {
+    @apply w-full border-collapse text-sm font-sans my-4 text-base-content;
+  }
+  thead {
+    @apply bg-base-200 text-base-content;
+  }
+  th {
+    @apply border border-base-300 px-4 py-2 text-left font-semibold;
+  }
+  td {
+    @apply border border-base-300 px-4 py-2 align-top;
+  }
+  tbody tr:nth-child(even) {
+    @apply bg-base-100;
+  }
+  tbody tr:hover {
+    @apply bg-base-200;
+  }
 }
 
 .btn-mini {