Gemma3 initial commit

bonsai/models/gemma3/README.md

@@ -0,0 +1,26 @@
+# Qwen3 in JAX
+
+This directory contains a pure JAX implementation of the [Gemma3 model](https://deepmind.google/models/gemma/gemma-3/), using the [Flax NNX](https://flax.readthedocs.io/en/v0.8.3/experimental/nnx/index.html) API. 
+
+Note that you need an access token to download the model weights. In order to run the scripts, make sure to save an environment variable `HF_TOKEN` with your huggingface access token. 
+
+
+## Model Configuration Support Status
+
+
+### Running this model
+
+
+```sh
+python3 -m bonsai.models.gemma3.tests.run_model
+```
+
+
+## How to contribute to this model
+
+### Remaining Tasks
+
+1. Implement with batching. Need this for FSDP. 
+2. Optimize based on the profiling. 
+3. Clean up code (variable names, etc.). Simplify unused configs (marked these with TODO) or use them.
+4. Update to include other model sizes and optimize parameter loading.