happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

convert : enable expert group selection for all models with it (#16691)

This commit is contained in:
Sigbjørn Skjæret 2025-10-26 17:21:23 +01:00 committed by GitHub
parent f696428ce8
commit 73a48c9790
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 8 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
convert_hf_to_gguf.py
View File

@ -742,6 +742,12 @@ class TextModel(ModelBase):
if (n_experts_used := self.hparams.get("num_experts_per_tok")) is not None: if (n_experts_used := self.hparams.get("num_experts_per_tok")) is not None:
self.gguf_writer.add_expert_used_count(n_experts_used) self.gguf_writer.add_expert_used_count(n_experts_used)
logger.info(f"gguf: experts used count = {n_experts_used}") logger.info(f"gguf: experts used count = {n_experts_used}")
if (n_expert_groups := self.hparams.get("n_group")) is not None:
self.gguf_writer.add_expert_group_count(n_expert_groups)
logger.info(f"gguf: expert groups count = {n_expert_groups}")
if (n_group_used := self.hparams.get("topk_group")) is not None:
self.gguf_writer.add_expert_group_used_count(n_group_used)
logger.info(f"gguf: expert groups used count = {n_group_used}")
if (head_dim := self.hparams.get("head_dim")) is not None: if (head_dim := self.hparams.get("head_dim")) is not None:
self.gguf_writer.add_key_length(head_dim) self.gguf_writer.add_key_length(head_dim)
@ -8233,8 +8239,6 @@ class BailingMoeV2Model(TextModel):
self.gguf_writer.add_expert_weights_scale(hparams["routed_scaling_factor"]) self.gguf_writer.add_expert_weights_scale(hparams["routed_scaling_factor"])
self.gguf_writer.add_expert_count(hparams["num_experts"]) self.gguf_writer.add_expert_count(hparams["num_experts"])
self.gguf_writer.add_expert_shared_count(hparams["num_shared_experts"]) self.gguf_writer.add_expert_shared_count(hparams["num_shared_experts"])
self.gguf_writer.add_expert_group_count(hparams["n_group"])
self.gguf_writer.add_expert_group_used_count(hparams["topk_group"])
self.gguf_writer.add_expert_weights_norm(hparams["norm_topk_prob"]) self.gguf_writer.add_expert_weights_norm(hparams["norm_topk_prob"])
if hparams["score_function"] == "sigmoid": if hparams["score_function"] == "sigmoid":

4
src/llama-model.cpp
View File

@ -6369,6 +6369,8 @@ void llama_model::print_info() const {
LLAMA_LOG_INFO("%s: n_ff = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_ff(il); }, hparams.n_layer).c_str()); LLAMA_LOG_INFO("%s: n_ff = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_ff(il); }, hparams.n_layer).c_str());
LLAMA_LOG_INFO("%s: n_expert = %u\n", __func__, hparams.n_expert); LLAMA_LOG_INFO("%s: n_expert = %u\n", __func__, hparams.n_expert);
LLAMA_LOG_INFO("%s: n_expert_used = %u\n", __func__, hparams.n_expert_used); LLAMA_LOG_INFO("%s: n_expert_used = %u\n", __func__, hparams.n_expert_used);
LLAMA_LOG_INFO("%s: n_expert_groups = %d\n", __func__, hparams.n_expert_groups);
LLAMA_LOG_INFO("%s: n_group_used = %d\n", __func__, hparams.n_group_used);
LLAMA_LOG_INFO("%s: causal attn = %d\n", __func__, hparams.causal_attn); LLAMA_LOG_INFO("%s: causal attn = %d\n", __func__, hparams.causal_attn);
LLAMA_LOG_INFO("%s: pooling type = %d\n", __func__, hparams.pooling_type); LLAMA_LOG_INFO("%s: pooling type = %d\n", __func__, hparams.pooling_type);
LLAMA_LOG_INFO("%s: rope type = %d\n", __func__, hparams.rope_type); LLAMA_LOG_INFO("%s: rope type = %d\n", __func__, hparams.rope_type);
@ -6469,8 +6471,6 @@ void llama_model::print_info() const {
LLAMA_LOG_INFO("%s: n_ff_exp = %d\n", __func__, hparams.n_ff_exp); LLAMA_LOG_INFO("%s: n_ff_exp = %d\n", __func__, hparams.n_ff_exp);
LLAMA_LOG_INFO("%s: n_ff_shexp = %d\n", __func__, hparams.n_ff_shexp); LLAMA_LOG_INFO("%s: n_ff_shexp = %d\n", __func__, hparams.n_ff_shexp);
LLAMA_LOG_INFO("%s: n_expert_shared = %d\n", __func__, hparams.n_expert_shared); LLAMA_LOG_INFO("%s: n_expert_shared = %d\n", __func__, hparams.n_expert_shared);
LLAMA_LOG_INFO("%s: n_expert_groups = %d\n", __func__, hparams.n_expert_groups);
LLAMA_LOG_INFO("%s: n_group_used = %d\n", __func__, hparams.n_group_used);
LLAMA_LOG_INFO("%s: expert_weights_scale = %.1f\n", __func__, hparams.expert_weights_scale); LLAMA_LOG_INFO("%s: expert_weights_scale = %.1f\n", __func__, hparams.expert_weights_scale);
LLAMA_LOG_INFO("%s: expert_weights_norm = %d\n", __func__, hparams.expert_weights_norm); LLAMA_LOG_INFO("%s: expert_weights_norm = %d\n", __func__, hparams.expert_weights_norm);
LLAMA_LOG_INFO("%s: expert_gating_func = %s\n", __func__, llama_expert_gating_func_name((llama_expert_gating_func_type) hparams.expert_gating_func)); LLAMA_LOG_INFO("%s: expert_gating_func = %s\n", __func__, llama_expert_gating_func_name((llama_expert_gating_func_type) hparams.expert_gating_func));