An entry in a Kaggle AI/ML competition to find extraterrestrial technosignatures in spectrogram data from deep space observations using radio telescopes.
An explanation of some non-standard arguments and shared volumes:
- Ports for Jupyter Notebook
- GPU Passthrough for CUDA-accelerated code
- Mount shm to increase shared memory, prevent out of memory efforts on large batch sizes
- Mount the poject repository
- Mount the data store
For the interactive situation, the run command is augmented with an interactive flag and an entrypoint to preempt the automated execution.
Running the docker container with the following command will cause the container to automatically train, save, and test the model detailed in seti_bl_pytorch_cnn.py
.
docker run -p 8888:8888 --gpus all --rm -v /dev/shm:/dev/shm -v /home/jeffrey/repos/SETI_Breakthrough_Listen:/SETI -v /home/jeffrey/data/seti_bl:/data setibl:0.1.0
To run the container and either work within the environment interactively or run the jupyter notebook, use the following command:
docker run -p 8888:8888 --gpus all --rm -v /dev/shm:/dev/shm -v /home/jeffrey/repos/SETI_Breakthrough_Listen:/SETI -v /home/jeffrey/data/seti_bl:/data -it --entrypoint bash setibl:0.1.0
And then at the container's bash prompt, use the following command to run the notebooks:
jupyter notebook --port=8888 --no-browser --ip=0.0.0.0 --allow-root
Any other python or tasks requireing the environment may be performed here as well.