- Accepted to NeurIPS 2020 Workshop: Self-Supervised Learning - Theory and Practice
- https://arxiv.org/abs/2103.08157
- Poster
- conda environment configs:
environment.yaml
- How to install:
conda env create -f environment.yaml
conda activate self-supervised-nas-3.6
pip install -e libs/nasbench --no-deps
pip install -e libs/NAS-Bench-201
CUDA_VISIBLE_DEVICES={device_index} python3 pretrain.py experiment=TripletMarginLoss
outputs/{date}/{time}/logs
-> tensorboard directory- Above directory can be checked with
tensorboard --bind_all --logdir=/path/to/log --port 8080
ortensorboard --bind_all --logdir_spec=exp1:/path/to/log1,exp2:/path/to/log2 --port 8080
outputs/{date}/{time}/weights
-> output model weights- Use with caution
pgrep -f "python pretrain.py" -a
will show all the training process.pkill -f "python pretrain.py"
will kill them all.
tmux ls
: Current tmux sessionstmux new -s {session_name}
: Create session withsession_name
tmux attach -t {session_name}
: Attach to existing session- (Inside tmux session)
Ctrl+a d
: Detach existing session - (Inside tmux session)
Ctrl+a PgUp/PgDn
: Scroll inside session (Ctrl+c
to abort) - (Inside tmux session)
Ctrl+d
: Close session