This is a merge of pre-trained language models created using
mergekit.
This model was merged using the breadcrumbs_ties merge method using
meta-llama/Llama-3.2-3B as a base.
1merge_method: breadcrumbs_ties
2base_model: meta-llama/Llama-3.2-3B
3tokenizer_source: PJMixers-Dev/LLaMa-3.2-Instruct-JankMix-v0.1-SFT-3B
4dtype: bfloat16
5parameters:
6 normalize: true
7models:
8 - model: meta-llama/Llama-3.2-3B-Instruct
9 parameters:
10 weight: 1
11 density: 0.9
12 gamma: 0.01
13 normalize: true
14 int8_mask: true
15 random_seed: 0
16 temperature: 0.5
17 top_p: 0.65
18 inference: true
19 max_tokens: 999999999
20 stream: true
21 quantization:
22 method: int8
23 value: 100
24 quantization:
25 method: int4
26 value: 100
27 - model: PJMixers-Dev/LLaMa-3.2-Instruct-JankMix-v0.1-SFT-3B
28 parameters:
29 weight: 1
30 density: 0.9
31 gamma: 0.01
32 normalize: true
33 int8_mask: true
34 random_seed: 0
35 temperature: 0.5
36 top_p: 0.65
37 inference: true
38 max_tokens: 999999999
39 stream: true
40 quantization:
41 method: int8
42 value: 100
43 quantization:
44 method: int4
45 value: 100
