Bonsai ShellAI: modified Gemma 4 E2B instruction-tuned derivative

Upstream model: google/gemma-4-E2B-it
Upstream authors: Google DeepMind
Source revision: 3e22461f65e89153144f8adb70e3b8c2cc9845a7
Source: https://huggingface.co/google/gemma-4-E2B-it/tree/3e22461f65e89153144f8adb70e3b8c2cc9845a7
Official license: https://ai.google.dev/gemma/apache_2

Modification notice: this distribution prunes MLP channels, exports the text
model to GGUF and applies mixed-precision quantization, including Q2_K per-layer
and tied main vocabulary embeddings. Details and measured limitations are in
README.md. This derivative is not an official Google DeepMind release.

The pinned upstream repository did not contain a separate NOTICE or LICENSE
file when inspected; its model card links the official Apache 2.0 terms.
This attribution/modification notice does not add license restrictions.