https://github.com/hesamsheikh/dataset_git_commands

https://github.com/hesamsheikh/dataset_git_commands

Last synced: 8 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/hesamsheikh/dataset_git_commands
• Owner: hesamsheikh
• Created: 2024-08-01T18:17:29.000Z (5 months ago)
• Default Branch: main
• Last Pushed: 2024-08-05T14:57:12.000Z (5 months ago)
• Last Synced: 2024-08-06T11:30:37.334Z (5 months ago)
• Language: Jupyter Notebook
• Size: 1.71 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Creating Synthetic Dataset Using Llama 3.1 405B and Nemotron 4

In this notebook we will use the following structure to create a synthetic dataset of Intructions and Git Commands.

![alt text](image.png)

We will create a set of instructions related to git queries in natural language, then we will generate the response for each instruction. 

The instruction/response pairs will be passed to a reward model, Nemotron 4, to filter out any bad pairs. 

Finally, the dataset will be pushed to HuggingFace. 

## Dataset

The final work can be viewed here:

https://huggingface.co/datasets/hesamsheikh/git-prompt