decentralised-ai / active-self-pretraining Goto Github PK

Run the following command to install the dependencies for the project:

pip install -r requirements.txt

Then, simply run for single GPU or CPU training:

python main.py

We used two datasets; ImageNet and Cifar10.

The datasets are available from this link: https://drive.google.com/drive/folders/1kIlQXoOl7tl8GwX-PKvvZR-Aklm16F59?usp=share_link

1. Create a folder named "dataset" in the project root folder
2. Download and extract the dataset(s) and then move them to the newly created directory.
3. Ensure you rename the ImageNet folder to "imagenet" and the Cifar10 folder to "cifar10v2"

You also need to create a folder named "save" in the project root folder where the checkpoints and other code generated files would be saved in.

$200 I will be carrying sufficiient clothing since I would be a 5 days trip.

Ask Prof. Crick if we need to pay for registration