Séb Arnold

Roviz: limit the number of visualized results

1

Until a more efficient visualization template is written, it is difficult to justify loading more than 1000-2000 experiments. So introduce the environment variable `ROVIZ_NRESULTS` which is equal to 1000 by...

Support snapshots

1

Snapshopts are unfinished experiments. The idea is to save partial results (in a dedicated folder) that could be used to implement fancier random search algorithms. (e.g. Population Based Training.) They...

Libraries we might want to cover: - [ ] PyTorch - [ ] TensorFlow - [ ] sklearn - [ ] MxNet - [ ] numpy/scipy

The idea is to execute SQL-like queries on the database of JSON files, for a given experiment. From the programmatic API, this means that you can get a list of...

Write tutorials

1

For now, they only contain guidelines on their content - [ ] Managing Experiments - [x] Optimizing Hyperparams - [x] Visualizing Results

Programmatic way of updating JSON summary instances

1

Since you can fetch a JSON summary, you should be able to modify some of its values, and either call `.save()` which creates a new result, or call `.update()` which...

GridSearch is implemented and merged in master, but needs unit tests.

Evolutionary is implemented and merged on master, but needs unit tests.

There should be an easy way to copy / remove results from one experiment to another, programmatically.

Add a button `Export Results` to save selected results in a zip file. All of this can be done in the client side using: https://github.com/Stuk/jszip