File size: 666 Bytes
fba85a6 b8d3525 fba85a6 b8d3525 6e2c2ed |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 |
---
{}
---
# miqu 70b
First model in the potential series.
## Prompt format: Mistral
```
<s> [INST] QUERY_1 [/INST] ANSWER_1</s> [INST] QUERY_2 [/INST] ANSWER_2</s>...
```
Beware that some backends (like llama.cpp) add bos already (by default), so you don't need to prepend it yourself.
## Settings
DO NOT CHANGE ROPE SETTINGS. This model uses high freq base with 32k seen tokens, it should be fine for most tasks.
Only tested with temp 1 and top_p 0.95 with everything else disabled.
<video src="https://cdn-uploads.huggingface.co/production/uploads/65ab93082bf3e0cbbf717850/cIEP5e43VP0k0caRzl16e.mp4" controls="controls" style="max-width: 720px;">
</video> |