nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 36 Packages Projects Releases Wiki Activity

model: support Ministral3 (#17644)

Browse Source 
* conversion script

* support ministral 3

* maybe this is better?

* add TODO for rope_yarn_log_mul

* better ppl (tested on 14B-Instruct)

* Add Ministral3 support to Mistral format

* improve arch handling

* add sizes

* Apply suggestions from code review

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

* nits

---------

Co-authored-by: Julien Denize <julien.denize@mistral.ai>
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
This commit is contained in:
Xuan-Son Nguyen
2025-12-01 12:26:52 +01:00
committed by GitHub
co-authored by Sigbjørn Skjæret Julien Denize
parent 649495c9d9
commit cd3c118908
11 changed files with 342 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/llama-hparams.h
+2 -2
View File
@@ -162,8 +162,8 @@ struct llama_hparams {
// llama4 smallthinker
uint32_t n_moe_layer_step = 0;
uint32_t n_no_rope_layer_step = 4;
uint32_t n_attn_temp_floor_scale = 8192;
float f_attn_temp_scale = 0.1;
uint32_t n_attn_temp_floor_scale = 0;
float f_attn_temp_scale = 0.0f;
// gemma3n altup
uint32_t n_altup = 4; // altup_num_inputs