32 lines
1.2 KiB
Markdown
32 lines
1.2 KiB
Markdown
|
|
||
|
|
---
|
||
|
|
|
||
|
|
library_name: transformers
|
||
|
|
base_model:
|
||
|
|
- meta-llama/Meta-Llama-3.1-8B-Instruct
|
||
|
|
datasets:
|
||
|
|
- jondurbin/airoboros-3.2
|
||
|
|
license: llama3
|
||
|
|
|
||
|
|
---
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
# QuantFactory/llama3.1-airoboros3.2-QDT-8B-GGUF
|
||
|
|
This is quantized version of [nbeerbower/llama3.1-airoboros3.2-QDT-8B](https://huggingface.co/nbeerbower/llama3.1-airoboros3.2-QDT-8B) created using llama.cpp
|
||
|
|
|
||
|
|
# Original Model Card
|
||
|
|
|
||
|
|
|
||
|
|
# llama3.1-airoboros3.2-QDT-8B
|
||
|
|
|
||
|
|
[meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) finetuned on [jondurbin/airoboros-3.2](https://huggingface.co/datasets/jondurbin/airoboros-3.2).
|
||
|
|
|
||
|
|
Llama 3 Instruct was used as the chat template.
|
||
|
|
|
||
|
|
### Method
|
||
|
|
|
||
|
|
Finetuned using a Google Colab A100 for 1 epoch. Only 25% of the data was used for training due to limited resources, hence QDT (Quarter Data Test).
|
||
|
|
|
||
|
|
[Based on mlabonne's SFT notebook.](https://colab.research.google.com/drive/1o_w0KastmEJNVwT5GoqMCciH-18ca5WS?usp=sharing)
|