proceduralia / experiment_buddy

mila_tools aims to reduce the overhead to deploy experiments on mila clusters.

experiment_buddy aims to reduce the overhead to deploy experiments on servers.

It's important to reduce cognitive overload for the researcher measured as seconds-to-first-tensorboard datapoint

Right now it's responsabilities cover:

1. Deployment on servers
2. Handling of Sweeps
3. Tracking of hyperparameters
4. Code versioning
5. Notifications
6. Wandb integration.

Example: (Updated)

1. python3 -m pip install virtualenv
2. python3 -m virtualenv venv --python=python3.8
3. source venv/bin/activate
4. Add cluster private-key to: https://github.com/settings/keys if you dont know this yet, check this out here
5. pip install wandb && wandb init there are two ways to set up wandb:
   • set up wandb init from the cmd
   • or set export WANDB_API_KEY = your wandb key which can be found here: https://wandb.ai/settings
6. python -m pip install jax jaxlib # Local requirements
7. python -m examples.mnist_classifier

More details on experiment-buddy:

1. experiment-buddy will commit to a branch called "experiment-buddy" to the first git repo found in the reverse os walk tree
2. Supports: Unix based OS