brucethemoose commited on
Commit
42550fc
1 Parent(s): 67f5f5a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +60 -2
README.md CHANGED
@@ -1,5 +1,63 @@
1
- ---
2
  license: other
3
  license_name: yi-license
4
- license_link: https://huggingface.co/01-ai/Yi-34B-200K/blob/main/LICENSE
 
 
 
 
5
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ --
2
  license: other
3
  license_name: yi-license
4
+ license_link: https://huggingface.co/01-ai/Yi-34B/blob/main/LICENSE
5
+ language:
6
+ - en
7
+ library_name: transformers
8
+ pipeline_tag: text-generation
9
  ---
10
+
11
+ **NousResearch/Nous-Capybara-34B** and **migtissera/Tess-M-v1.2** merged with an expermental implementation of "dare" via mergekit. See:
12
+
13
+ > Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch
14
+
15
+ https://arxiv.org/abs/2311.03099
16
+
17
+ https://github.com/cg123/mergekit/tree/dare-tokenizer
18
+
19
+ Merged with the following config, and the tokenizer from Yi Llamafied:
20
+
21
+ ```
22
+ models:
23
+ - model: /home/alpha/Storage/Models/Raw/larryvrh_Yi-34B-200K-Llamafied
24
+ # no parameters necessary for base model
25
+ - model: /home/alpha/Storage/Models/Raw/migtissera_Tess-M-v1.2
26
+ parameters:
27
+ weight: 0.62
28
+ density: 0.55
29
+ - model: /home/alpha/Storage/Models/Raw/Nous-Capybara-34B
30
+ parameters:
31
+ weight: 0.56
32
+ density: 0.55
33
+ merge_method: dare_ties
34
+ base_model: /home/alpha/Storage/Models/Raw/larryvrh_Yi-34B-200K-Llamafied
35
+ parameters:
36
+ int8_mask: true
37
+ dtype: bfloat16
38
+
39
+ ```
40
+
41
+ ## Prompt template: Orca-Vicuna
42
+
43
+ ```
44
+ SYSTEM: {system_message}
45
+ USER: {prompt}
46
+ ASSISTANT:
47
+
48
+ ```
49
+ Sometimes the model "spells out" the stop token as `</s>` like Capybara, so you may need to add `</s>` as an additional stopping condition.
50
+
51
+ ***
52
+
53
+ Credits:
54
+
55
+ https://github.com/cg123/mergekit/tree/dare-tokenizer
56
+
57
+ https://huggingface.co/NousResearch/Nous-Capybara-34B/
58
+
59
+ https://huggingface.co/migtissera/Tess-M-v1.2
60
+
61
+ https://huggingface.co/larryvrh/Yi-34B-200K-Llamafied
62
+
63
+ https://huggingface.co/01-ai/Yi-34B-200K