EELML Gemma 4 E2B IT · MLX Q4
EELML Studio runtime-ready package compiled from
google/gemma-4-E2B-it.
This repository distributes the EELML single-file physical package, not the
original checkpoint. The package is pre-quantized and pre-arranged for the
dedicated mlx_gemma4_e execution path. EELML Runtime performs package
verification, read-only mapping, zero-copy binding, and compute; it does not
load or transform source weights at runtime.
Package contract
| Field | Value |
|---|---|
| File | gemma4-e2b-it-mlx-q4-v1.eelml |
| Package ID | gemma4-e2b-it-mlx-q4-v1 |
| Canonical graph | gemma4_e.canonical@1 |
| Physical layout | mlx_gemma4_e_q4_v1 |
| Runtime | eelml-mlx-gemma4-e 3.x |
| Protocol adapter | gemma4-native-tool-v1@1 |
| Chat template | gemma4-turn-v1 |
| Size | 2,636,692,297 bytes |
| SHA-256 | ffdf60db912753d07e3d17f7926b8ba470518a9a9d8f064ad87829643310445c |
Installation
Install this model from EELML Studio's model manager. The App downloads the
package, verifies its internal manifest and SHA-256, then transactionally
records the model/runtime/protocol binding in SQLite. Do not rename another
weight file to .eelml or select this package with a different runtime.
Scope
The current EELML package contains the text-generation path used by EELML Studio. It does not advertise the upstream model's image, audio, or video input paths.
License and attribution
The upstream model is published by Google DeepMind under Apache License 2.0. The EELML physical package remains subject to the upstream model license and attribution requirements.
这是为 EELML Studio 专用 mlx_gemma4_e Runtime 深度编译的单文件模型包。
客户端只通过模型管理器下载、校验和安装；模型包、Runtime、协议适配器与聊天模板任一身份不一致时必须拒绝加载。