40 lines
1.2 KiB
Markdown
40 lines
1.2 KiB
Markdown
---
|
|
tags:
|
|
- text-generation-inference
|
|
- transformers
|
|
- unsloth
|
|
- llama
|
|
datasets:
|
|
- CohereLabs/aya_dataset
|
|
- databricks/databricks-dolly-15k
|
|
- OpenAssistant/oasst1
|
|
- HuggingFaceH4/no_robots
|
|
language:
|
|
- en
|
|
base_model:
|
|
- common-pile/comma-v0.1-2t
|
|
- dianab1729/Ada-1
|
|
---
|
|
|
|
# Model Card for Comma-QLoRa-IT
|
|
|
|
QLora instruction fine-tuning on Common-Pile's Comma-v0.1-2T model. The goal is to make a usable assistant while maintaining the goal of not using stolen data.
|
|
|
|
dianab1729/Ada-1 (which was already SFT'd on no-robots) was used as a base to skip a step.
|
|
|
|
- SFT on The human authored, english part from CohereLabs/aya_dataset
|
|
- SFT on databricks/databricks-dolly-15k
|
|
- DPO on OpenAssistant/oasst1
|
|
|
|
All QLoRA.
|
|
It ain't much, but it's honest work. :)
|
|
|
|
Planning on making a much better version of this, I'm just learning on this rn.
|
|
|
|
# Uploaded finetuned model
|
|
- **Finetuned from model :** dianab1729/Ada-1
|
|
|
|
This llama model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
|
|
|
|
[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)
|