gemma-quickstart

Quickstarter for Gemma by Google

Gemma Types

1. 1B
   Light weight text model, ideal for small applications
2. 4B
   Balanced for performance and flexibility, with multimodal support
3. 12B
   Strong language capabilities, designed for complex tasks
4. 27B Enhanced understanding, great for sophesticated applications

Finetuning

1. Keras

Serving

1. Llama.cpp
2. Keras

Interface

1. LMStudio
2. Keras

References

1. Demo: Rapid prototyping with Gemma and Llama.cpp
2. Demo: Building a Gemma chatbot with Keras
   2.1 Discusses on finetuning with LORA
3. Demo: Gemma on-device with MediaPipe and TensorFlow Lite
4. Demo: JAX, Flax and Gemma
5. Run Gemma with Keras