Image-Text-to-Text
MLX
Safetensors
unlimited-ocr
ax-engine
mlx-vlm
ocr
mxfp8
int8
apple-silicon
automatosx
conversational
8-bit precision
Instructions to use AutomatosX/AX-Unlimited-OCR-3B-MoE-MLX-MXFP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use AutomatosX/AX-Unlimited-OCR-3B-MoE-MLX-MXFP8 with MLX:
# Make sure mlx-vlm is installed # pip install --upgrade mlx-vlm from mlx_vlm import load, generate from mlx_vlm.prompt_utils import apply_chat_template from mlx_vlm.utils import load_config # Load the model model, processor = load("AutomatosX/AX-Unlimited-OCR-3B-MoE-MLX-MXFP8") config = load_config("AutomatosX/AX-Unlimited-OCR-3B-MoE-MLX-MXFP8") # Prepare input image = ["http://images.cocodataset.org/val2017/000000039769.jpg"] prompt = "Describe this image." # Apply chat template formatted_prompt = apply_chat_template( processor, config, prompt, num_images=1 ) # Generate output output = generate(model, processor, formatted_prompt, image) print(output) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Document Hub release tags v0.1.0 and v0.2.0
Browse files
README.md
CHANGED
|
@@ -16,6 +16,25 @@ widget:
|
|
| 16 |
- text: "<image>Free OCR. Extract all text."
|
| 17 |
---
|
| 18 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
# AX Unlimited-OCR 3B MoE — MLX (MXFP8 + affine INT8 head)
|
| 20 |
|
| 21 |
This is an AutomatosX sensitivity-guided MLX conversion of
|
|
|
|
| 16 |
- text: "<image>Free OCR. Extract all text."
|
| 17 |
---
|
| 18 |
|
| 19 |
+
## Release tags
|
| 20 |
+
|
| 21 |
+
| Tag | Commit | Notes |
|
| 22 |
+
|-----|--------|-------|
|
| 23 |
+
| **v0.2.0** (current) | `6708e4d86b0a342c5891e47516aa412bbfd2eda2` | Sensitivity-guided MXFP8 + calibrated affine INT8 LM head |
|
| 24 |
+
| **v0.1.0** | `fd0c04fc20db886f8a80fe1177081b07206586c7` | First OCR-aware MXFP8 publish |
|
| 25 |
+
|
| 26 |
+
Pin production loads with `revision="v0.2.0"` (or the commit SHA). Floating `main` may move on future publishes.
|
| 27 |
+
|
| 28 |
+
```bash
|
| 29 |
+
python -m mlx_vlm.generate \
|
| 30 |
+
--model AutomatosX/AX-Unlimited-OCR-3B-MoE-MLX-MXFP8 \
|
| 31 |
+
--revision v0.2.0 \
|
| 32 |
+
--image document.png \
|
| 33 |
+
--prompt '<image>document parsing.' \
|
| 34 |
+
--max-tokens 4096
|
| 35 |
+
```
|
| 36 |
+
|
| 37 |
+
|
| 38 |
# AX Unlimited-OCR 3B MoE — MLX (MXFP8 + affine INT8 head)
|
| 39 |
|
| 40 |
This is an AutomatosX sensitivity-guided MLX conversion of
|