model: tag ffn_latent as MUL_MAT to fix buft probe (#23664)

ffn_latent_down/up are declared GGML_OP_MUL in LLM_TENSOR_INFOS but
nemotron-h feeds them through ggml_mul_mat. The loader buft probe asks
the backend about the declared op, so it tested an elementwise MUL on a
q8_0 weight. That used to return true unconditionally and the weight
stayed on GPU by luck. Once supports_op told the truth, the probe got a
no and the loader pushed the weight and its matmul to CPU, splitting the
graph. Tagging it MUL_MAT asks the real question, the math is unchanged.

Verified on Nemotron 3 Super 120B Q5_K_M: from 64.9 back to 103.22 t/s.

diff --git a/src/llama-arch.cpp b/src/llama-arch.cpp
index c9eead18aa39158f1882dd8c139198681ec00665..e7d027c981451409f209ad1d436296e570a6462c 100644 (file)
--- a/src/llama-arch.cpp
+++ b/src/llama-arch.cpp
@@ -767,8 +767,9 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
     {LLM_TENSOR_NEXTN_SHARED_HEAD_HEAD,     {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_NEXTN_SHARED_HEAD_NORM,     {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
     // Nemotron 3 Super
-    {LLM_TENSOR_FFN_LATENT_DOWN,            {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
-    {LLM_TENSOR_FFN_LATENT_UP,              {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL}},
+    // latent projections feed ggml_mul_mat, the buft probe must use MUL_MAT to keep them on GPU
+    {LLM_TENSOR_FFN_LATENT_DOWN,            {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
+    {LLM_TENSOR_FFN_LATENT_UP,              {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
 };
 
 LLM_KV::LLM_KV(llm_arch arch, const char * suffix) : arch(arch), suffix(suffix) {}