Skip to content

Commit d02130d

Browse files
committed
llama : print DeekSeek-V2-specific parameters in llm_load_print_meta()
1 parent 5cc7ec1 commit d02130d

File tree

1 file changed

+10
-0
lines changed

1 file changed

+10
-0
lines changed

llama.cpp

+10
Original file line numberDiff line numberDiff line change
@@ -4989,6 +4989,16 @@ static void llm_load_print_meta(llama_model_loader & ml, llama_model & model) {
49894989
if (vocab.special_suffix_id != -1) { LLAMA_LOG_INFO( "%s: SUF token = %d '%s'\n", __func__, vocab.special_suffix_id, vocab.id_to_token[vocab.special_suffix_id].text.c_str() ); }
49904990
if (vocab.special_middle_id != -1) { LLAMA_LOG_INFO( "%s: MID token = %d '%s'\n", __func__, vocab.special_middle_id, vocab.id_to_token[vocab.special_middle_id].text.c_str() ); }
49914991
if (vocab.special_eot_id != -1) { LLAMA_LOG_INFO( "%s: EOT token = %d '%s'\n", __func__, vocab.special_eot_id, vocab.id_to_token[vocab.special_eot_id].text.c_str() ); }
4992+
4993+
if (model.arch == LLM_ARCH_DEEPSEEK2) {
4994+
LLAMA_LOG_INFO("%s: n_layer_dense_lead = %d\n", __func__, hparams.n_layer_dense_lead);
4995+
LLAMA_LOG_INFO("%s: n_lora_q = %d\n", __func__, hparams.n_lora_q);
4996+
LLAMA_LOG_INFO("%s: n_lora_kv = %d\n", __func__, hparams.n_lora_kv);
4997+
LLAMA_LOG_INFO("%s: n_ff_exp = %d\n", __func__, hparams.n_ff_exp);
4998+
LLAMA_LOG_INFO("%s: n_expert_shared = %d\n", __func__, hparams.n_expert_shared);
4999+
LLAMA_LOG_INFO("%s: expert_weights_scale = %.1f\n", __func__, hparams.expert_weights_scale);
5000+
LLAMA_LOG_INFO("%s: rope_yarn_log_mul = %.4f\n", __func__, hparams.rope_yarn_log_mul);
5001+
}
49925002
}
49935003

49945004
// Returns false if cancelled by progress_callback

0 commit comments

Comments
 (0)