v0.2.10

Peg MLX to 0.0.14 in all plugins

v0.2.9

In this release, we force conda to be in a specific location

After a lot of testing, we have found that this is the only reliable solution for working with conda because of how electron shell and how conda interact. We now install conda to a fixed place (~/miniconda3) and only use the conda that is installed there, manually referring to it each time we need it. We do not automatically install conda to the users shell -- we manually inject the conda init script every time we need to use conda

v0.2.8.2

upgrade dependencies when installing them

v0.2.8.1

update requirements

v0.2.8

• Change default iterations on mlx-trainer
• Offer data that can be used by the app to help with plugin updates
• Fix large issue with mlx_lora_trainer
• Much cleaner, more reliable install script

v0.2.7

Patch release to add tensorboard logging to MLX training, improve plugin management, plus a variety of bug fixes.

New Contributors

• @ai-made-approachable made their first contribution in #6

Full Changelog: v0.2.6...v0.2.7

v0.2.6

Fixed a number of issues training models using mlx_lora_trainer.

Full Changelog: v0.2.5...v0.2.6

v0.2.5

Patch release to fix two issues:

• training job looking in wrong directory when trying to run locally created datasets
• training output log file was concatenating on one single file for all training jobs

Full Changelog: v0.2.4...v0.2.5

v0.2.4

Patch release to fix a number of issues in the datasets API, mostly around error handling and reporting.

Full Changelog: v0.2.3...v0.2.4

v0.2.3

Patch release to fix downloading of a number of models.

Full Changelog: v0.2.2...v0.2.3