Deacon-34b-4bit / README.md
KnutJaegersberg's picture
Update README.md
7e8a506
|
raw
history blame
1.18 kB
metadata
license: other
license_name: yi-license
license_link: LICENSE
datasets:
  - totally-not-an-llm/EverythingLM-data-V3
pipeline_tag: text-generation

image/png

The perfect organism.

An adapter for KnutJaegersberg/Yi-34B-Llamafied. 5 epochs with NEFTune. The brand new bitsandbytes 4-bit file format.

Run pip install -U bitsandbytes and install this pull requested transformer version: https://github.com/poedator/transformers/tree/save4

Might still be buggy. It was a bit too early. I'll wait until the functionality is officially supported, due to some cuda issues. I don't want to resolve those, I'll just wait a couple of days.

Prompt Example:

### System:

You are an AI assistant. User will give you a task. Your goal is to complete the task as faithfully as you can. While performing the task think step-by-step and justify your steps.

### Instruction: 

How do you fine tune a large language model? 

### Response:

image/png