Skip to content

LLM inference in C/C++

C++ ggml

dflash : clarify output logging of target_layer_ids (#27013)

Daniel Bevenius2bacf9ea5cf9
diff --git a/src/models/dflash.cpp b/src/models/dflash.cpp
index eb36336762..daaa20826a 100644
--- a/src/models/dflash.cpp
+++ b/src/models/dflash.cpp
@@ -14,11 +14,14 @@ void llama_model_dflash::load_arch_hparams(llama_model_loader & ml) {
hparams.n_embd_inp_enc_impl = (uint32_t) target_layer_ids.size() * hparams.n_embd;
- LLAMA_LOG_INFO("%s: DFlash extract_layers = [", __func__);
- for (size_t i = 0; i < target_layer_ids.size(); ++i) {
- LLAMA_LOG_INFO("%d%s", target_layer_ids[i], i + 1 < target_layer_ids.size() ? ", " : "");
+ std::string layers;
+ const char * sep = "";
+ for (const auto id : target_layer_ids) {
+ layers += sep;
+ layers += std::to_string(id);
+ sep = ", ";
}
- LLAMA_LOG_INFO("]\n");
+ LLAMA_LOG_INFO("%s: DFlash extract_layers = [%s]\n", __func__, layers.c_str());
// DeepSeek-V4 DSpark backbone: stages are full DSV4 blocks, uniform sliding window (the draft KV ring)
ml.get_key(LLM_KV_HYPER_CONNECTION_COUNT, hparams.dsv4_hc_mult, false);