16 lines (12 loc) · 907 Bytes

Instructions for running on Caltech GPUs

Log onto https://imperium-sm.hep.caltech.edu and start your server
Start a new terminal by clicking on the top right

Clone our DeepJet repository as usual, but do not install miniconda2

git clone https://github.com/DeepDoubleB/DeepJet

Go back to the directory tree (original tab) and navigate to the DeepJet/Train folder

To run a training, open and run the notebook DeepJet/Train/train_deepdoubleb_Resnet.py. You can evaluate each cell of the notebook by doing shift+enter once your cursor is in that cell.
To evalate the performance after the training, open the notebook DeepJet/Train/train_deepdoubleb_Resnet.py and execute each cell of the notebook (Note this training is not working correctly yet... but the plots are still made).