labmlai / app Goto Github PK

Show only the latest session for a computer in the list view. Show the previous sessions for the same computer inside the header view.

Currently, the server only uses data from the master process in a distributed run. ideally, all the processes should be aggregated.

Right now, it shows a blank screen. Show some sort of a error instead of the blank screen

Runs list page & computers list page doesn't update automatically, if coming back from another page, when the cache is already present.

Service needs to be implemented to communicate the server and a local machine.

• Sync deleted/active/stale runs.
• Delete checkpoints/artefacts/Tensor board logs/runs from UI.
• Send disk space usage of each runs to the server.
• Check whether the local machine is actively communicating with the server.

This would give a seamless experience on page reload after initial login

It will be great, if you add Dockerfile and docker-compose.yaml files to repo for easy installation of self hosted servers

1. Currently, some of the hyperparameters are showing incorrect (rounded integer) values. GET /api/v1/run/[id] is returning correct types and values.
2. Hyperparameter saving is not working for all hyperparameters (no matter if it is displayed correctly), either by tweaking sliders or by manually entering values. POST /api/v1/hyper_params/[id] is returning 500 now.

Though these are more likely two separate issues, they both appeared today (on app.labml.ai).

The screenshot of the first issue:

The numbers on the Run page shows correct numbers:

Currently, we use linear scale smoothing for the Log-chart. Preferably, Log scale smoothing should be used.

So, the user can configure what should be plotted in the graph by default

y axis of the charts are slightly off aligned