Views
No views yet
1dtype: bfloat16
2tokenizer:
3 source: union
4merge_method: ffn_kv_injection
5base_model: meta-llama/Meta-Llama-3-8B-Instruct
6models:
7- model: meta-llama/Meta-Llama-3-8B-Instruct
8 parameters: {}
9- model: shenzhi-wang/Llama3-8B-Chinese-Chat
10 parameters: {}
11parameters:
12 weights:
13 - 0.7
14 - 0.3
15 tau: 1.0
16 aspect_thresh: 1.5
17 pnorm: 2.0
18 alpha_floor: 0.0
19 alpha_ceil: 1.0
20write_readme: README.md
21