This is a merge of pre-trained language models created using
mergekit.
This model was merged using the
DARE TIES merge method using
MihaiPopa-1/LFM2.5-350M-heretic as a base.
1merge_method: dare_ties
2base_model: MihaiPopa-1/LFM2.5-350M-heretic
3dtype: bfloat16
4
5parameters:
6 normalize: false
7 int8_mask: true
8
9models:
10 # slot_1_strongest
11 - model: mkurman/LiquidAI-LFM2.5-350M-SYNTH
12 parameters:
13 weight: 0.28
14 density: 0.38
15
16 # slot_2_medium
17 - model: appvoid/palmer-005-core
18 parameters:
19 weight: 0.22
20 density: 0.34
21
22 # slot_3_light
23 - model: squ11z1/claude-oss-350m
24 parameters:
25 weight: 0.14
26 density: 0.28
27
