---
datasets:
- gozfarb/ShareGPT_Vicuna_unfiltered
---

## LoRA Info:
Please note that this is a highly experimental LoRA model. It may do some good stuff, it might do some undesirable stuff. Hopefully after training is fully done, it'll be for sure good!

Current upload: checkpoint of epoch 2.47 (of 3)

The alpaca-lora repo by tloen doesn't *seem* to create an immediately working adapter/peft model until the training is finalized, but I attempted to turn a checkpoint into one.
I hope it works for y'all ^~^ Thanks to MetaIX for initial seemingly successful testing of the first uploaded checkpoint (epoch 0.8) as well as epoch 1.

**Wikitext2 eval benchmark for epoch 0.8 upload (4-bit merge):** 4.548329830169678 (Thanks to Metal/MetaIX)