Exclude project_per_layer_input by matching node names

This ensures that all other graphs which don't exhibit this pattern do
not have their behavior changed.

Author: ORippler

ggml/src/ggml-cuda/ggml-cuda.cu

@@ -2589,7 +2589,9 @@ static bool check_node_graph_compatibility_and_refresh_copy_ops(ggml_backend_cud
 
     // Loop over nodes in GGML graph to obtain info needed for CUDA graph
     cuda_ctx->cuda_graph->cpy_dest_ptrs.clear();
-    std::uint8_t batch_size_counter = 0;
+
+    const std::string gemma3n_per_layer_proj_src0_name = "inp_per_layer_selected";
+    const std::string gemma3n_per_layer_proj_src1_name = "per_layer_proj";
 
     for (int i = 0; i < cgraph->n_nodes; i++) {
         ggml_tensor * node = cgraph->nodes[i];
@@ -2612,19 +2614,16 @@ static bool check_node_graph_compatibility_and_refresh_copy_ops(ggml_backend_cud
 #endif
         }
 
-        if (node->op == GGML_OP_ADD && node->src[1] && node->src[1]->ne[1] > 1) {
-            // disable CUDA graphs for batch size > 1 for now. The heuristic here allows to use CUDA graphs
-            // for Gemma3n, which uses a single Matrix-Matrix Addition as part of `project_per_layer_input`, while detecting
-            // batched execution for all graphs with >1 GGML_OP_ADD nodes. See also
+        if (node->op == GGML_OP_ADD && node->src[1] && node->src[1]->ne[1] > 1 && (node->src[0] ? node->src[0]->name != gemma3n_per_layer_proj_src0_name : true) && (node->src[1] ? node->src[1]->name != gemma3n_per_layer_proj_src1_name : true)) {
+            // disable CUDA graphs for batch size > 1 for now while excluding the matrix-matrix addition as part of Gemma3n's `project_per_layer_input` operation
+            // by means of matching node names. See
+            // https://github.com/ggml-org/llama.cpp/blob/f9a31eea06a859e34cecb88b4d020c7f03d86cc4/src/llama-model.cpp#L10199-L10241 and
             // https://github.com/huggingface/transformers/blob/bda75b4011239d065de84aa3e744b67ebfa7b245/src/transformers/models/gemma3n/modeling_gemma3n.py#L1773,
             // Generally, changes in batch size or context size can cause changes to the grid size of some kernels.
-            ++batch_size_counter;
-            if (batch_size_counter > 1) {
-                use_cuda_graph = false;
-                #ifndef NDEBUG
-                  GGML_LOG_DEBUG("%s: disabling CUDA graphs due to repeated batch size > 1 [%s] [%ld %ld %ld %ld]\n", __func__, node->name, node->ne[0], node->ne[1], node->ne[2], node->ne[3]);
-                #endif
-            }
+            use_cuda_graph = false;
+            #ifndef NDEBUG
+                GGML_LOG_DEBUG("%s: disabling CUDA graphs due to batch size > 1 [%s] [%ld %ld %ld %ld]\n", __func__, node->name, node->ne[0], node->ne[1], node->ne[2], node->ne[3]);
+            #endif
         }
 
         if (node->op == GGML_OP_CPY) {