thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

graph : remove redundant scale_w parameter (#20235)

Browse source
This commit is contained in:
Sigbjørn Skjæret 2026-03-08 18:58:28 +01:00 • committed by GitHub
parent 451ef08432
commit 35bee031e1
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
41 changed files with 85 additions and 86 deletions
Download patch file Download diff file Expand all files Collapse all files

4
src/models/bailingmoe2.cpp
View file

@ -1,7 +1,5 @@
#include "models.h"
llm_build_bailingmoe2::llm_build_bailingmoe2(const llama_model & model, const llm_graph_params & params) :
llm_graph_context(params) {
const int64_t n_embd_head = hparams.n_embd_head_v;
@ -90,7 +88,7 @@ llm_build_bailingmoe2::llm_build_bailingmoe2(const llama_model & model, const ll
model.layers[il].ffn_exp_probs_b,
n_expert, n_expert_used,
LLM_FFN_SILU, hparams.expert_weights_norm,
hparams.expert_weights_scale, hparams.expert_weights_scale,
hparams.expert_weights_scale,
(llama_expert_gating_func_type) hparams.expert_gating_func,
il);
cb(moe_out, "ffn_moe_out", il);