Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
base_model:
|
4 |
+
- mistralai/Mistral-Nemo-Instruct-2407
|
5 |
+
---
|
6 |
+
|
7 |
+
# Sto'Vo'Kor 12B
|
8 |
+
|
9 |
+
[mistralai/Mistral-Nemo-Instruct-2407](https://huggingface.co/mistralai/Mistral-Nemo-Instruct-2407) fine-tuned on a private collection of ~30M tokens worth of instruct and multi-turn RP.
|
10 |
+
|
11 |
+
## Instruct Format
|
12 |
+
|
13 |
+
Instruct format is V3-Tekken, the same as Mistral Nemo Instruct (except the chat template used won't freak out if your turns get mixed up, like tends to happen in ST. Thanks, fizz!)
|
14 |
+
|
15 |
+
```
|
16 |
+
<s>[INST]{System or user instructions}[/INST]{AI Response}</s>
|
17 |
+
```
|
18 |
+
During training, system turns were given as the first user turn in the conversation, separate from the user character's first turn. i.e., System as user -> AI turn (filler or first turn) -> User first turn
|
19 |
+
|
20 |
+
## Recommended Samplers
|
21 |
+
|
22 |
+
Whatever you're used to for Nemo should work. For me this is stable with:
|
23 |
+
- temp 0.7
|
24 |
+
- min-p 0.03
|
25 |
+
- DRY 0.5/1.75/5/1024
|