Distributed Parallel Training for MHCCL Experiment logs: Dataset: Epilepsy for test Train only for 3 epochs. Pre-train: 1 node with 4 GPU; Time for one epoch: 178.95 seconds Evaluate: Using 1 GPU.
Running commands: Pretrain: python main_sally.py --dataset_name epilepsy --epochs 100 --seed 0 --dist_url 'tcp://localhost:10001' --multiprocessing_distributed --world_size 1 --rank 0
Evaluate: python classifier.py --dataset_name epilepsy --pretrained experiment_dataset_name/checkpoint_epochs_seed.pth.tar --lr 5 --seed 0 --dist_url 'tcp://localhost:10001' --multiprocessing_distributed --world_size 1 --rank 0 --id epilepsy_linear_0001
git add . git commit -m "Initial commit" git push -u origin master