Misc. fixes.

Author: pwilkin

src/llama-arch.cpp

@@ -2546,6 +2546,7 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
     {LLM_TENSOR_SSM_X,                      {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_SSM_DT,                     {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_SSM_OUT,                    {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
+    {LLM_TENSOR_SSM_BETA_ALPHA,             {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_TIME_MIX_W1,                {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_TIME_MIX_W2,                {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_TIME_MIX_A1,                {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},

src/llama-context.cpp

@@ -1388,7 +1388,7 @@ void llama_context::output_reorder() {
 
 uint32_t llama_context::graph_max_nodes() const {
     if (model.arch == LLM_ARCH_QWEN3NEXT) {
-        return std::max<uint32_t>(8192, 32u*model.n_tensors());
+        return std::max<uint32_t>(8192u, 32u*model.n_tensors());
     }
     return std::max<uint32_t>(1024u, 8u*model.n_tensors());
 }