Skip to content

LLM inference in C/C++

C++ ggml

model : fix gemma4-assistant (#28183)

Sigbjørn Skjæret73159c30399a
diff --git a/src/llama-model.cpp b/src/llama-model.cpp
index 6c79971ced..bfce09de0c 100644
--- a/src/llama-model.cpp
+++ b/src/llama-model.cpp
@@ -1224,7 +1224,7 @@ void llama_model_base::load_hparams(llama_model_loader & ml) {
ml.get_key(LLM_KV_BLOCK_COUNT, hparams.n_layer_all);
GGML_ASSERT(hparams.n_layer_all > 0 && hparams.n_layer_all <= LLAMA_MAX_LAYERS);
ml.get_key(LLM_KV_NEXTN_PREDICT_LAYERS, hparams.n_layer_nextn, false);
- GGML_ASSERT(hparams.n_layer_nextn < hparams.n_layer_all);
+ GGML_ASSERT(hparams.n_layer_nextn <= hparams.n_layer_all);
ml.get_key(LLM_KV_EXPERT_COUNT, hparams.n_expert, false);
ml.get_key(LLM_KV_EXPERT_USED_COUNT, hparams.n_expert_used, false);
ml.get_key(LLM_KV_EXPERT_GROUP_COUNT, hparams.n_expert_groups, false);
diff --git a/src/models/gemma4-assistant.cpp b/src/models/gemma4-assistant.cpp
index 989fa42c85..8431ec2a1f 100644
--- a/src/models/gemma4-assistant.cpp
+++ b/src/models/gemma4-assistant.cpp
@@ -4,7 +4,7 @@ void llama_model_gemma4_assistant::load_arch_hparams(llama_model_loader & ml) {
hparams.n_embd_inp_impl = hparams.n_embd_out();
hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
- ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer());
+ ml.get_key_or_arr(LLM_KV_ATTENTION_SLIDING_WINDOW_PATTERN, hparams.is_swa_impl, hparams.n_layer_all);
uint32_t n_kv_shared_layers = 0;
ml.get_key(LLM_KV_ATTENTION_SHARED_KV_LAYERS, n_kv_shared_layers, false);