From: Alessandro98-git <redacted>
Date: Tue, 23 Dec 2025 02:04:57 +0000 (+0100)
Subject: model : fix div-by-zero for Nemotron V2 (#18309)
X-Git-Tag: upstream/0.0.7599~83
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=96e33a814e100025648d12eb6a827154b48277d9;p=pkg%2Fggml%2Fsources%2Fllama.cpp

model : fix div-by-zero for Nemotron V2 (#18309)

* llama-model : fix Nemotron V2 crash by moving MoE parameters calculation

* remove whitespace

---------

Co-authored-by: Sigbjørn Skjæret <redacted>
---

diff --git a/src/llama-model.cpp b/src/llama-model.cpp
index 87fefd57..0d5bcc64 100644
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -5243,9 +5243,6 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                     const int64_t n_group    = hparams.ssm_n_group;
                     const int64_t d_in_proj  = 2*d_inner + 2*n_group*d_state + n_ssm_head;
 
-                    const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
-                    const int64_t n_ff_shexp = hparams.n_ff_shexp;
-
                     // embeddings
                     tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
 
@@ -5297,6 +5294,9 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                             layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias",   i), {n_embd},         TENSOR_NOT_REQUIRED);
                         }  else {
                             if (n_expert != 0) {
+                                const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
+                                const int64_t n_ff_shexp = hparams.n_ff_shexp;
+
                                 layer.ffn_gate_inp    = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), { n_embd, n_expert}, 0);
                                 layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert         }, 0);