Results and Explanation

For a more detailied look at the project and see who collaborated, read the PDF file

Run single-GPU

main.py contains the code for single-GPU (or CPU) operation. The EA can be enabled, or a manual genotype can be used for full training

Run multi-GPU

NOTE: The multi-GPU code has NOT been tested recently due to unavailability of multiple GPUs on the HPC

Run the following command from the root directory to run the multi-GPU training script:

torchrun --nnodes=1 --nproc-per-node=2 ./train_multi_gpu.py

If you want to use more than 2 GPUs, change the --nproc-per-node=<n> parameter.

Notes

• NOTE: Change the GPU numbers before running the script to prevent hijacking someone else's GPU!
• The single GPU training file does not work as of now (unless necessary in the future, I will not fix that).