Datasets
Models
5 lines (4 with data), 429 Bytes
- First, you have to download the thought viz dataset pickle file and its equivalent images (images are not required for feature extraction learning, just class names are required).
- Then run train.py code and it will run for around 3000 epochs.
- After train, now run the find_bestckpt.py code, it will give you the path of best checkpoint with better test acc.
- Finally you can run inference.py for generating tsne plot.