What's Changed
- fix rope when very long sequence precision is key by @vince62s in #200
- Better fix for long rope (training was not optimized) by @vince62s in #201
- add filtertooshort transform by @vince62s in #202
- Basic pixtral support, paving the way for vision models 🖼️ by @francoishernandez in #153
- Clean / rename / simplify by @vince62s in #203
- Bump 0.1.1 by @francoishernandez in #205
Full Changelog: 0.1.0...0.1.1
Contributors
francoishernandez and vince62s