19 lines
598 B
Markdown
19 lines
598 B
Markdown
|
## Model Info
|
||
|
|
Merge of my [VicUnlocked-alpaca-half-30b LoRA](https://huggingface.co/Aeala/VicUnlocked-alpaca-half-30b-LoRA)
|
||
|
|
|
||
|
|
**Important Note**: While this is trained on a cleaned ShareGPT dataset like Vicuna used, this was trained in the *Alpaca* format, so prompting should be something like:
|
||
|
|
|
||
|
|
```
|
||
|
|
### Instruction:
|
||
|
|
|
||
|
|
<prompt> (without the <>)
|
||
|
|
|
||
|
|
### Response:
|
||
|
|
```
|
||
|
|
|
||
|
|
## Benchmarks
|
||
|
|
wikitext2: 4.372413635253906
|
||
|
|
ptb-new: 24.69171714782715
|
||
|
|
c4-new: 6.469308853149414
|
||
|
|
|
||
|
|
Results generated with GPTQ evals (not quantized) thanks to [Neko-Institute-of-Science](https://huggingface.co/Neko-Institute-of-Science)
|