cip-pytorch-minimal

A set of minimal tutorials for setting up and using Pytorch on the CIP computers at LMU

Consider reading some of the suggested reads

The aim is to collect here worked out tutorials and use cases for methods we are actively looking to apply for cosmological data

Initial tests to include:

• GAN
• cGAN
• Normalizing flows

Where to run

LMU machines

CIP GPU server (login needed with token) you can reserve Nvidia A40 gpus for model training

Alternatively at cip-sv-login0[1234].cip.physik.uni-muenchen.de

There's also a cpu only server, that however does not provide full torch support CIP CPU server (login needed with token)

acccess to LMU CIP

You'll need to enroll a loging token at https://otp.physik.uni-muenchen.de in addition to the physik account

before running

Be sure to reserve a GPU first, torch install works only when cuda gpu is available

• run sh pip.sh This should perform multiple downloads, not, then something is not going right