Alex Cheema
Alex Cheema
In the dashboard, there is the download page which shows all the models downloaded onto each node in your cluster. It would be great to extend this to not just...
Currently we have an environment variable to control the model directory (see https://github.com/exo-explore/exo/blob/main/src/exo/shared/constants.py). However, there's no way to set this via the macOS app. It would be great if this...
Things like model directory are set once via environment variables currently. If you want to change them, you need to restart the app with different environment variables. It would be...
## Describe the bug In `mlx_distributed_init` we create a devices file (in the case of `MlxJaccl`) or a hosts file (in the case of `MlxRing`). The name we use for...
A blind spot for testing is installation / bootstrapping for a normal user. For example, a user ran into an issue with RDMA because they didn't have RDMA enabled #1004...
## Describe the bug If you go to the dashboard initially it will either show models as 0% downloaded or 100% downloaded, i.e. it doesn't show any partial downloads. ##...
Create a Known Issues section in the README, and add the Local Network Access toggle fix there.
## Motivation We added a download page to the dashboard which shows the currently download status of each model on each node. Users have reported this to be extremely useful....
## Describe the bug After restarting EXO, nodes no longer discover each other over TB5 and therefore do not show any RDMA placement previews. Lots of logs from RUST, possibly...
## Describe the bug Deleting an instance, creating a new one, then sends a chat completion request with the old, stale instance. ## To Reproduce Steps to reproduce the behavior:...