LightTuning
π€ CLI to fine-tune GPT-3.5 Turbo models rapidly. π¬
π Concept
LightTuning is a quick way to fine-tune GPT-3.5 Turbo models. To scale and enhance a model from a small conversation to a light GPT-3.5 model that can overperform GPT-4. How cool is that?
π§βπ» Usage
Installation
Install the required packages using pip.
pip3 install -r requirements.txt
Run the main.py
script to start the tool.
python3 main.py
You will be greeted with a welcome message and a list of actions to choose from:
- Generate dataset
- Upload dataset
- Fine-tune model
- All
Select an action using the arrow keys and press Enter.
Generate Dataset
If you choose 'Generate dataset', you will be asked to provide the path to the seed conversation. If you don't provide a path, LightTuning will ask you to enter it.
Add a seed conversation to the file. The conversation should be in the following OpenAI conversation format:
[
{
"role": "system",
"content": "You are an assistant that can translate from English to Spanish."
},
{
"role": "user",
"content": "Hello"
},
{
"role": "assistant",
"content": "Hola"
},
{
"role": "user",
"content": "Happy"
},
{
"role": "assistant",
"content": "Feliz"
}
]
Upload Dataset
If you choose 'Upload dataset', you will be asked to enter the path to the dataset file. If you don't provide a path, the tool will ask you to enter it. You will then be asked to confirm the upload.
Fine-tune Model
If you choose 'Fine-tune model', you will be asked to provide the file ID for fine-tuning. If you don't provide a file ID, the tool will ask you to enter it.
All
If you choose 'All', the tool will execute all of the above actions in order.
π License
Distributed under the MIT License.
See LICENSE
for more information.