To setup my code on windows (note that only the training code works on windows):
cd <project_path>
conda create env -f enviroment/anaconda_enviroments/ml_compression_enviroment.yml
To setup my code on ubuntu (this works for both training and testing):
cd <project_path>
conda create env -f enviroment/anaconda_enviroments/ubuntu_env.yml
To run a training session (for one epoch):
-s is the directory to save the model
-i is the path to the imagenet dataset
-e is the number of epochs
-b is the batch size
-m is the embedding dimension
-t is the transfer dimension
-w is the window size
-d is the depth of the network
python session.py -s ./saved_models -i <imagenet_path> -e 1 -b 8 -m 24 -t 16 -w 2 -d 3
To test the compression (for one image):
-m is the model directory
-s is a random seed
-i is the path to the imagenet dataset
-n is the number of images to be compressed
python test_compression.py -m ./saved_models/<model_name> -i <imagenet_path> -s 55 -n 1
To test quantisation:
-m is the model directory
-i is the path to the imagenet dataset
python test_quantisation.py -m ./saved_model/<model_name> -i <imagenet_path>