search

noahgolmant / pytorch-hessian-eigenthings

Efficient PyTorch Hessian eigendecomposition tools!
MIT License
360 stars 43 forks source link

Python Error: the following arguments are required: experimentname #39

Closed AjinkyaBankar closed 3 years ago

AjinkyaBankar commented 3 years ago

Hello Sir, When I am executing pytorch-hessian-eigenthings/example/main.py I get the following error:

usage: main.py [-h] [--self_host SELF_HOST] [--cpu] [--port PORT]
               [--server_port SERVER_PORT] [--config CONFIG]
               [--devices_per_trial DEVICES_PER_TRIAL] [--dataroot DATAROOT]
               [--s3 S3] [--logroot LOGROOT] [--seed SEED]
               [--num_eigenthings NUM_EIGENTHINGS] [--batch_size BATCH_SIZE]
               [--eval_batch_size EVAL_BATCH_SIZE] [--momentum MOMENTUM]
               [--num_steps NUM_STEPS] [--max_samples MAX_SAMPLES] [--cuda]
               [--full_dataset] [--fname FNAME] [--mode {power_iter,lanczos}]
               experimentname
main.py: error: the following arguments are required: experimentname
An exception has occurred, use %tb to see the full traceback. 

SystemExit: 2

What should be the experimentname while executing the code? Kindly help. Thanks in advance.

noahgolmant commented 3 years ago

The example file uses skeletor to track experiment data. I will clarify that in the documentation or (more likely) remove that dependency. But the experimentname can be any meaningful name for your experiment, such as resnet18_cifar.