jukofyork commited on
Commit
5571b53
1 Parent(s): 08b8aec

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +9 -1
README.md CHANGED
@@ -14,7 +14,7 @@ Other experimental models which attempt to encourage more diverse/creative text
14
  - [creative-writer-v0.1-bravo-35b](https://huggingface.co/jukofyork/creative-writer-v0.1-bravo-35b) - Scaled the pre-softmax logits by `1.1` during training (and then reset after training).
15
  - [creative-writer-v0.1-charlie-35b](https://huggingface.co/jukofyork/creative-writer-v0.1-charlie-35b) - Scaled the pre-softmax logits by `0.9` during training (and didn't reset after training).
16
 
17
- <details> <summary>Click to see some (brief) tests on effect of these changes</summary>
18
 
19
  #### Using `command-r-3-2024` with `temperature = 1` and `min-p = 0.01`:
20
 
@@ -76,6 +76,14 @@ Other experimental models which attempt to encourage more diverse/creative text
76
 
77
  ![image.png](https://cdn-uploads.huggingface.co/production/uploads/65995c45539c808e84c38bf1/PwnDkctZ273zMHC-Ta_YU.png)
78
 
 
 
 
 
 
 
 
 
79
  </details>
80
 
81
  ---
 
14
  - [creative-writer-v0.1-bravo-35b](https://huggingface.co/jukofyork/creative-writer-v0.1-bravo-35b) - Scaled the pre-softmax logits by `1.1` during training (and then reset after training).
15
  - [creative-writer-v0.1-charlie-35b](https://huggingface.co/jukofyork/creative-writer-v0.1-charlie-35b) - Scaled the pre-softmax logits by `0.9` during training (and didn't reset after training).
16
 
17
+ <details> <summary>Click to see some (brief) tests on the effect of these changes</summary>
18
 
19
  #### Using `command-r-3-2024` with `temperature = 1` and `min-p = 0.01`:
20
 
 
76
 
77
  ![image.png](https://cdn-uploads.huggingface.co/production/uploads/65995c45539c808e84c38bf1/PwnDkctZ273zMHC-Ta_YU.png)
78
 
79
+ ---
80
+
81
+ **Observations**:
82
+
83
+ - Up-scaling of the pre-softmax logits during training used by `creative-writer-v0.1-bravo:35b` looks the most promising.
84
+ - Down-scaling of the pre-softmax logits during training used by `creative-writer-v0.1-charlie:35b` looks to be very similar to inference-time temperature adjustment.
85
+ - It may be better to just leave the pre-softmax logits up-scaled after training and then let the user perform inference-time temperature adjustment.
86
+
87
  </details>
88
 
89
  ---