Update README.md
Browse files
README.md
CHANGED
|
@@ -2,13 +2,22 @@
|
|
| 2 |
license: llama2
|
| 3 |
---
|
| 4 |
CodeLlama 2 7b
|
|
|
|
| 5 |
With Guanaco Lora (Tim Dettmers), merged by Varunk29.
|
|
|
|
| 6 |
Then
|
|
|
|
| 7 |
With Mistral AI 7b 0.1 delta bits compared to Llama2 (extracted by Undi95), merged by me.
|
| 8 |
|
|
|
|
|
|
|
| 9 |
Base model (CodeLlama) training context : 16k (max context up to 96k with the base ROPE)
|
|
|
|
| 10 |
Mistral injection training context : 8k (Sliding Windows Attention is likely inoperant on such a merge/injection)
|
| 11 |
|
|
|
|
|
|
|
| 12 |
For test and amusement only.
|
| 13 |
|
|
|
|
| 14 |
Prompt : Alpaca works.
|
|
|
|
| 2 |
license: llama2
|
| 3 |
---
|
| 4 |
CodeLlama 2 7b
|
| 5 |
+
|
| 6 |
With Guanaco Lora (Tim Dettmers), merged by Varunk29.
|
| 7 |
+
|
| 8 |
Then
|
| 9 |
+
|
| 10 |
With Mistral AI 7b 0.1 delta bits compared to Llama2 (extracted by Undi95), merged by me.
|
| 11 |
|
| 12 |
+
---
|
| 13 |
+
|
| 14 |
Base model (CodeLlama) training context : 16k (max context up to 96k with the base ROPE)
|
| 15 |
+
|
| 16 |
Mistral injection training context : 8k (Sliding Windows Attention is likely inoperant on such a merge/injection)
|
| 17 |
|
| 18 |
+
---
|
| 19 |
+
|
| 20 |
For test and amusement only.
|
| 21 |
|
| 22 |
+
|
| 23 |
Prompt : Alpaca works.
|