Instructions to use mnmly/rfdetr-small-mlx-fp32 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use mnmly/rfdetr-small-mlx-fp32 with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir rfdetr-small-mlx-fp32 mnmly/rfdetr-small-mlx-fp32
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
| { | |
| "model_type": "rf-detr", | |
| "encoder": "dinov2_windowed_small", | |
| "hidden_dim": 256, | |
| "resolution": 512, | |
| "dec_layers": 3, | |
| "num_queries": 300, | |
| "num_classes": 90, | |
| "patch_size": 16, | |
| "num_windows": 2, | |
| "group_detr": 13, | |
| "sa_nheads": 8, | |
| "ca_nheads": 16, | |
| "dec_n_points": 2, | |
| "two_stage": true, | |
| "bbox_reparam": true, | |
| "lite_refpoint_refine": true, | |
| "layer_norm": true, | |
| "out_feature_indexes": [ | |
| 3, | |
| 6, | |
| 9, | |
| 12 | |
| ], | |
| "projector_scale": [ | |
| "P4" | |
| ], | |
| "positional_encoding_size": 32 | |
| } |