Instructions to use carlesonielfa/MinerU2.5-Pro-2604-1.2B-mlx-8bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use carlesonielfa/MinerU2.5-Pro-2604-1.2B-mlx-8bit with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download carlesonielfa/MinerU2.5-Pro-2604-1.2B-mlx-8bit --local-dir MinerU2.5-Pro-2604-1.2B-mlx-8bit
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
MinerU2.5-Pro-2604-1.2B MLX 8bit
8-bit quantized MLX conversion of opendatalab/MinerU2.5-Pro-2604-1.2B for Apple Silicon.
- Quantization: 8-bit, group size 64
- Format: MLX SafeTensors
- Disk size: ~1.7 GB
- Framework: mlx-vlm
Conversion
python -m mlx_vlm convert \
--hf-path opendatalab/MinerU2.5-Pro-2604-1.2B \
--mlx-path MinerU2.5-Pro-2604-1.2B-mlx-8bit \
-q --q-bits 8 --q-group-size 64
Note: A
tie_word_embeddingsbug inmlx_vlm'sqwen2_vlconfig parser was patched locally during conversion (nestedtext_configkeys were being discarded, causing a missinglm_head.weight).
Important: Image Size
The layout pass requires images resized to 1036ร1036 before inference. Passing images at their original resolution causes the model to produce malformed output (<|box_start|>1.1.1.1... loop). The content pass (per-block crops) works at natural crop dimensions.
Related Models
- bf16:
carlesonielfa/MinerU2.5-Pro-2604-1.2B-mlx-bf16 - 8-bit (this model):
carlesonielfa/MinerU2.5-Pro-2604-1.2B-mlx-8bit - 4-bit:
carlesonielfa/MinerU2.5-Pro-2604-1.2B-mlx-4bit - Original: opendatalab/MinerU2.5-Pro-2604-1.2B
License
Inherits the license from the original model.
- Downloads last month
- 24
Model size
1B params
Tensor type
U32
ยท
BF16 ยท
Hardware compatibility
Log In to add your hardware
8-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐ Ask for provider support
Model tree for carlesonielfa/MinerU2.5-Pro-2604-1.2B-mlx-8bit
Base model
opendatalab/MinerU2.5-Pro-2604-1.2B