v0.2.0 - Gemma 2 2b, Compilation support, Assisted Genertion

This release follows the advent of Gemma 2 2b, the smallest Gemma 2 model available to date. Its small size is perfect for limited hardware resources or as a speed-up companion to the larger variants through assisted generation!

Highlights:

• 💎 Gemma 2 2b support
• ⚡ speed preset (uses torch.compile)
• 👪 CLI uses the 2b to speed up the larger models, through assisted generation

Thank you to all contributors since the previous release: @sanchit-gandhi @gante @shapito27 @Vaibhavs10

v0.1.0 - Local Gemma is born

Release v0.1.0

Local Gemma is born! 💎

In this release, we include:

• Simple installation instructions
• Automatic parameter inference, based on the hardware available at call time
• A minimal CLI
• A Python package, so you can ship your app and ensure the best flags are used on the user side