Dracones commited on
Commit
970d431
1 Parent(s): 453bfff

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +5 -2
README.md CHANGED
@@ -41,6 +41,7 @@ Below are the perplexity scores for the EXL2 models. A lower score is better.
41
  | Quant Level | Perplexity Score | Repo |
42
  |-------------|------------------|------|
43
  | 6.0 | 4.7068 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
 
44
  | 5.0 | 4.7309 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
45
  | 4.5 | 4.8111 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
46
  | 4.25 | 4.8292 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
@@ -49,9 +50,10 @@ Below are the perplexity scores for the EXL2 models. A lower score is better.
49
  | 3.5 | 4.9592 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
50
  | 3.25 | 5.0631 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
51
  | 3.0 | 5.2050 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
 
52
  | 2.5 | 5.6681 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
 
53
 
54
- _TODO:_ Dracones quant perplexity scores being worked on.
55
 
56
  ## EQ Bench
57
 
@@ -60,10 +62,11 @@ Here are the EQ Bench scores for the EXL2 quants using Alpaca, ChatML, Command-R
60
  | Quant Size | Alpaca | ChatML | Command-R | Command-R-Plus |
61
  |------------|--------|--------|--------|--------|
62
  | 6.0 | 70.77 | 62.58 | 75.81 | 74.95 |
 
63
  | 5.0 | 69.51 | 63.94 | 74.92 | 75.28 |
64
 
65
 
66
- _TODO:_ EQ Bench scores being worked on.
67
 
68
  ### Command-R-Plus Template
69
 
 
41
  | Quant Level | Perplexity Score | Repo |
42
  |-------------|------------------|------|
43
  | 6.0 | 4.7068 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
44
+ | 5.5 | 4.7136 | Dracones |
45
  | 5.0 | 4.7309 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
46
  | 4.5 | 4.8111 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
47
  | 4.25 | 4.8292 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
 
50
  | 3.5 | 4.9592 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
51
  | 3.25 | 5.0631 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
52
  | 3.0 | 5.2050 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
53
+ | 2.75 | 5.3820 | Dracones |
54
  | 2.5 | 5.6681 | [turboderp](https://huggingface.co/turboderp/command-r-plus-103B-exl2) |
55
+ | 2.25 | 5.9769 | Dracones |
56
 
 
57
 
58
  ## EQ Bench
59
 
 
62
  | Quant Size | Alpaca | ChatML | Command-R | Command-R-Plus |
63
  |------------|--------|--------|--------|--------|
64
  | 6.0 | 70.77 | 62.58 | 75.81 | 74.95 |
65
+ | 5.5 | 71.93 | 67.7 | 74.9 | 75.48 |
66
  | 5.0 | 69.51 | 63.94 | 74.92 | 75.28 |
67
 
68
 
69
+ _Note:_ EQ Bench scripting not working well, other quants may not be tested.
70
 
71
  ### Command-R-Plus Template
72