Dracones commited on
Commit
74f8473
1 Parent(s): d78c85a

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +10 -10
README.md CHANGED
@@ -41,36 +41,36 @@ Below are the perplexity scores for the EXL2 models. A lower score is better.
41
 
42
  ## EQ Bench
43
 
44
- Here are the EQ Bench scores for the EXL2 quants using Alpaca, ChatML and Vicuna-Chat prompt templates. A higher score is better.
45
 
46
 
47
  | Quant Size | Instruct Template | Score |
48
  |------------|-------------------|-------|
49
  | 8.0 | Alpaca | 75.02 |
50
  | 8.0 | ChatML | 73.82 |
51
- | 8.0 | Vicuna-Chat | 73.53 |
52
  | 7.0 | Alpaca | 74.96 |
53
  | 7.0 | ChatML | 74.22 |
54
- | 7.0 | Vicuna-Chat | 72.71 |
55
  | 6.0 | Alpaca | 75.11 |
56
  | 6.0 | ChatML | 74.29 |
57
- | 6.0 | Vicuna-Chat | 74.16 |
58
  | 5.0 | Alpaca | 74.11 |
59
  | 5.0 | ChatML | 72.74 |
60
- | 5.0 | Vicuna-Chat | 73.28 |
61
  | 4.5 | Alpaca | 73.18 |
62
  | 4.5 | ChatML | 73.62 |
63
- | 4.5 | Vicuna-Chat | 73.32 |
64
  | 4.0 | Alpaca | 71.85 |
65
  | 4.0 | ChatML | 73.99 |
66
- | 4.0 | Vicuna-Chat | 74.74 |
67
  | 3.5 | Alpaca | 73.73 |
68
  | 3.5 | ChatML | 72.07 |
69
- | 3.5 | Vicuna-Chat | 73.56 |
70
 
71
- ### Vicuna Chat Template
72
 
73
- This is the Vicuna-Chat template yaml that was used in EQ bench. It was tested in Text Generation Web UI and seemed to produce accurate results.
74
 
75
  ```yaml
76
  user: "USER:"
 
41
 
42
  ## EQ Bench
43
 
44
+ Here are the EQ Bench scores for the EXL2 quants using Alpaca, ChatML and Chat-Vicuna prompt templates. A higher score is better.
45
 
46
 
47
  | Quant Size | Instruct Template | Score |
48
  |------------|-------------------|-------|
49
  | 8.0 | Alpaca | 75.02 |
50
  | 8.0 | ChatML | 73.82 |
51
+ | 8.0 | Chat-Vicuna | 73.53 |
52
  | 7.0 | Alpaca | 74.96 |
53
  | 7.0 | ChatML | 74.22 |
54
+ | 7.0 | Chat-Vicuna | 72.71 |
55
  | 6.0 | Alpaca | 75.11 |
56
  | 6.0 | ChatML | 74.29 |
57
+ | 6.0 | Chat-Vicuna | 74.16 |
58
  | 5.0 | Alpaca | 74.11 |
59
  | 5.0 | ChatML | 72.74 |
60
+ | 5.0 | Chat-Vicuna | 73.28 |
61
  | 4.5 | Alpaca | 73.18 |
62
  | 4.5 | ChatML | 73.62 |
63
+ | 4.5 | Chat-Vicuna | 73.32 |
64
  | 4.0 | Alpaca | 71.85 |
65
  | 4.0 | ChatML | 73.99 |
66
+ | 4.0 | Chat-Vicuna | 74.74 |
67
  | 3.5 | Alpaca | 73.73 |
68
  | 3.5 | ChatML | 72.07 |
69
+ | 3.5 | Chat-Vicuna | 73.56 |
70
 
71
+ ### Chat-Vicuna Template
72
 
73
+ This is the Chat-Vicuna template yaml that was used in EQ bench. It was tested in Text Generation Web UI and seemed to produce accurate results.
74
 
75
  ```yaml
76
  user: "USER:"