Update README.md
Browse files
README.md
CHANGED
|
@@ -4,6 +4,7 @@ datasets:
|
|
| 4 |
- neph1/bellman-7b-finetune
|
| 5 |
language:
|
| 6 |
- sv
|
|
|
|
| 7 |
---
|
| 8 |
|
| 9 |
Qlora trained for 5 epochs on 6400 rows of q&a from around 1000 pages from wikipedia + around 100 of python questions and examples from
|
|
@@ -11,6 +12,14 @@ eph1/Alpaca-Lora-GPT4-Swedish-Refined (because I had spent so much time cleaning
|
|
| 11 |
gathered examples and some generated using chat-gpt.
|
| 12 |
Dataset otherwise generated using gpt-3.5-turbo.
|
| 13 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
I may run another 5 epochs on this. But it feels like it's 'aligned' pretty well. (Regular mistral insists Magdalena Andersson is prime minister, still.)
|
| 15 |
|
| 16 |
Example (q8):
|
|
|
|
| 4 |
- neph1/bellman-7b-finetune
|
| 5 |
language:
|
| 6 |
- sv
|
| 7 |
+
library_name: peft
|
| 8 |
---
|
| 9 |
|
| 10 |
Qlora trained for 5 epochs on 6400 rows of q&a from around 1000 pages from wikipedia + around 100 of python questions and examples from
|
|
|
|
| 12 |
gathered examples and some generated using chat-gpt.
|
| 13 |
Dataset otherwise generated using gpt-3.5-turbo.
|
| 14 |
|
| 15 |
+
Rank: 16
|
| 16 |
+
|
| 17 |
+
Alpha: 16
|
| 18 |
+
|
| 19 |
+
Dropout: 0.1
|
| 20 |
+
|
| 21 |
+
Context length: 1024
|
| 22 |
+
|
| 23 |
I may run another 5 epochs on this. But it feels like it's 'aligned' pretty well. (Regular mistral insists Magdalena Andersson is prime minister, still.)
|
| 24 |
|
| 25 |
Example (q8):
|