Commit
•
ef1b582
1
Parent(s):
0d26c60
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: mistralai/Mistral-Nemo-Base-2407
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
datasets:
|
| 5 |
+
- BeaverAI/Nemo-Inst-Tune-ds
|
| 6 |
+
language:
|
| 7 |
+
- en
|
| 8 |
+
library_name: transformers
|
| 9 |
+
---
|
| 10 |
+
# Dory 12b (v2)
|
| 11 |
+
(redone) redone instruct finetune of mistral nemo 12b's base. *not* (E)RP-focused, leave that to drummer.
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
|
| 15 |
+
thanks to twisted again for the compute :3
|
| 16 |
+
|
| 17 |
+
## Prompting
|
| 18 |
+
alpaca-like:
|
| 19 |
+
```
|
| 20 |
+
### System:
|
| 21 |
+
[Optional system prompt]
|
| 22 |
+
|
| 23 |
+
### Instruction:
|
| 24 |
+
[Query]
|
| 25 |
+
|
| 26 |
+
### Response:
|
| 27 |
+
[Response]</s>
|
| 28 |
+
|
| 29 |
+
### Instruction:
|
| 30 |
+
[...]
|
| 31 |
+
```
|
| 32 |
+
|
| 33 |
+
## Training details
|
| 34 |
+
Rank 64 QDoRA, trained on the following data mix:
|
| 35 |
+
- All of [kalomaze/Opus_Instruct_3k](https://huggingface.co/datasets/kalomaze/Opus_Instruct_3k)
|
| 36 |
+
- All conversations with a reward model rating above 5 in [Magpie-Align/Magpie-Gemma2-Pro-Preview-Filtered](https://huggingface.co/datasets/Magpie-Align/Magpie-Gemma2-Pro-Preview-Filtered)
|
| 37 |
+
- 50k of [Gryphe/Sonnet3.5-SlimOrcaDedupCleaned](https://huggingface.co/datasets/Gryphe/Sonnet3.5-SlimOrcaDedupCleaned)
|
| 38 |
+
- All stories above 4.7 rating and published before 2020 in [Fizzarolli/FallingThroughTheSkies-592k-Filtered-Filtered](https://huggingface.co/datasets/Fizzarolli/FallingThroughTheSkies-592k-Filtered-Filtered)
|