jeiku commited on
Commit
16e2c2b
1 Parent(s): f8b9d97

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +10 -0
README.md CHANGED
@@ -6,6 +6,16 @@ library_name: transformers
6
  ---
7
  ![image/png](https://cdn-uploads.huggingface.co/production/uploads/626dfb8786671a29c715f8a9/elC-5Dq32CK2lOemM9gd2.png)
8
 
 
 
 
 
 
 
 
 
 
 
9
  This model was created with the help of several members of Anthracite.
10
 
11
  This is a 4B parameter Minitron derivative healed and instruct tuned on 70M high quality tokens. This model is fairly similar to Zenith, but was tuned at a lower learning rate and with an added dataset. This model was tuned at 8k context during all steps. This model should perform well as a general assistant and RP model.
 
6
  ---
7
  ![image/png](https://cdn-uploads.huggingface.co/production/uploads/626dfb8786671a29c715f8a9/elC-5Dq32CK2lOemM9gd2.png)
8
 
9
+ ```
10
+ | Groups |Version|Filter|n-shot|Metric| |Value | |Stderr|
11
+ |------------------|------:|------|------|------|---|-----:|---|-----:|
12
+ |mmlu | 2|none | |acc |↑ |0.5903|± |0.0039|
13
+ | - humanities | 2|none | |acc |↑ |0.5481|± |0.0068|
14
+ | - other | 2|none | |acc |↑ |0.6601|± |0.0082|
15
+ | - social sciences| 2|none | |acc |↑ |0.6786|± |0.0082|
16
+ | - stem | 2|none | |acc |↑ |0.4983|± |0.0086|
17
+ ```
18
+
19
  This model was created with the help of several members of Anthracite.
20
 
21
  This is a 4B parameter Minitron derivative healed and instruct tuned on 70M high quality tokens. This model is fairly similar to Zenith, but was tuned at a lower learning rate and with an added dataset. This model was tuned at 8k context during all steps. This model should perform well as a general assistant and RP model.