Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: gemma
|
| 3 |
+
base_model: google/gemma-4-31B-it-qat-q4_0-unquantized
|
| 4 |
+
tags:
|
| 5 |
+
- apertura
|
| 6 |
+
- apml
|
| 7 |
+
- mlx
|
| 8 |
+
- quantized
|
| 9 |
+
- gemma4
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# gemma-4-31b-it-qat-q4-g32.apml — Apertura model bundle
|
| 13 |
+
|
| 14 |
+
Gemma 4 31B instruction-tuned, Google's QAT-Q4 weights converted to the
|
| 15 |
+
**Apertura `.apml` bundle format**: MLX-affine 4-bit quantization
|
| 16 |
+
(group size 32, 8-bit embeddings), `tokenizer.json` +
|
| 17 |
+
`chat_template.jinja` included, runtime `mlx`, architecture `gemma4`.
|
| 18 |
+
|
| 19 |
+
Consumed by [Apertura](https://github.com/apocryphx/Apertura) — an
|
| 20 |
+
Objective-C++ transformer engine + macOS chat app on MLX.
|
| 21 |
+
|
| 22 |
+
## Download
|
| 23 |
+
|
| 24 |
+
```sh
|
| 25 |
+
hf download apocryphx/gemma-4-31b-it-qat-q4-g32-apml --local-dir gemma-4-31b-it-qat-q4-g32.apml
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## License
|
| 29 |
+
|
| 30 |
+
Gemma is provided under and subject to the
|
| 31 |
+
[Gemma Terms of Use](https://ai.google.dev/gemma/terms). By downloading this
|
| 32 |
+
model you agree to those terms, including the
|
| 33 |
+
[Gemma Prohibited Use Policy](https://ai.google.dev/gemma/prohibited_use_policy).
|
| 34 |
+
This repository redistributes a quantized **Model Derivative** of
|
| 35 |
+
`google/gemma-4-31B-it-qat-q4_0-unquantized`; the same terms and use
|
| 36 |
+
restrictions apply to it and to any further derivatives.
|