v0.10.2

What's Changed

• Docker support for CUDA GPUs by @vasanthkumar-in in #51
• Docker update by @vasanthkumar-in in #50
• Change ollama installer to create a link in the models directory. by @dadmobile in #63
• Fix/importer better errors by @dadmobile in #64
• Run ruff format all of API by @dadmobile in #67
• fix recipe names and ids to work with the new way tensorboard does stuff by @sanjaycal in #68
• Create the first dummy Fast API test by @aliasaria in #65
• Remove unneeded requirements and update readme by @sanjaycal in #66
• Update Delete HF Model from Cache by @abhimazu in #62
• Adding Weights and Biases support for all training plugins by @deep1401 in #69
• Bump the pip group with 6 updates by @dependabot in #70
• Ollama plugin: Support for variants and improved naming by @dadmobile in #71

New Contributors

• @vasanthkumar-in made their first contribution in #51
• @dependabot made their first contribution in #70

Full Changelog: v0.10.1...v0.10.2

v0.10.1

What's Changed

• Add Unsloth Grpo plugin by @sanjaycal in #55
• Cancel model download by @abhimazu in #56
• Fix bug which caused document preview issues within folders by @deep1401 in #57
• Enable multiple runs to be shown in one tensorboard for the same task by @deep1401 in #58

New Contributors

• @abhimazu made their first contribution in #56

Full Changelog: v0.10.0...v0.10.1

v0.10.0

• ollama plugin
• document generation support
• documents directory support
• detailed reports for evals
• python completion recipe
• bug fixes for adapters and installer dependencies

v0.9.7

Patch update to switch from pip to uv for python library installation.

v0.9.6

Patch release with big updates to evals and a few bug fixes:

• RAG: openai Connection Error on new installs
• MLX: always show adapters on Foundation page

v0.9.5

Patch update to bump the version of flash_attn to work properly on clean installs.

v0.9.4

Full Changelog: v0.9.3...v0.9.4

v0.9.3

Full Changelog: v0.9.2...v0.9.3

v0.9.2

What's Changed

• Fix evals deletion by @deep1401 in #38
• Fix issues for MLX LoRA for making model fusing optional as per #42 on app repo by @deep1401 in #33
• Pin llama cpp by @sanjaycal in #35
• Creating a new plugin for MLX and removing Fastchat dependency for evals by @deep1401 in #39
• Adding DeepEval evaluation by @deep1401 in #41
• Add bitsandbytes to llama trainer by @sanjaycal in #42
• Adding GEval support by @deep1401 in #43

Full Changelog: v0.9.1...v0.9.2

v0.9.1

What's Changed

• fix recipes and add batch size by @sanjaycal in #36
• Add default batch size to avoid failing by @sanjaycal in #37

Full Changelog: v0.9.0...v0.9.1