Instructions to use mlx-community/Hi3Dv1.0-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use mlx-community/Hi3Dv1.0-mlx with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download mlx-community/Hi3Dv1.0-mlx --local-dir Hi3Dv1.0-mlx
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Download validation.json from mlx-community/Hi3Dv1.0-mlx: direct link, hf CLI and curl.
- Browser
- Download file 3.87 kB
-
https://huggingface.co/mlx-community/Hi3Dv1.0-mlx/resolve/main/validation.json
- Command line
-
hf download hf://mlx-community/Hi3Dv1.0-mlx/validation.json
-
curl -L -o validation.json https://huggingface.co/mlx-community/Hi3Dv1.0-mlx/resolve/main/validation.json
3.87 kB
| { | |
| "mlx_version": "0.32.4", | |
| "tensor_count": 1899, | |
| "parameter_elements": 4312697600, | |
| "weight_bytes": 10158060544, | |
| "dtypes": { | |
| "float32": 1151, | |
| "float16": 748 | |
| }, | |
| "strict_load": { | |
| "missing": 0, | |
| "extra": 0, | |
| "shape_mismatches": 0 | |
| }, | |
| "regression_tests": { | |
| "class": "TestHi3D", | |
| "passed": 14 | |
| }, | |
| "reference_comparison": { | |
| "reference": "Original Hi3D PyTorch code (github 23dcc68, hub 705fc09) on CPU; spconv and flash-attn replaced by reference implementations; float32 unless noted", | |
| "mlx_device": "GPU", | |
| "reports": { | |
| "sphere64_report": { | |
| "conditioning": { | |
| "max_abs": 5.714595317840576e-06, | |
| "mean_abs": 2.006490236685449e-07, | |
| "rel_l2": 8.875883576616418e-06 | |
| }, | |
| "sigmas": { | |
| "max_abs": 5.960464477539063e-08, | |
| "mean_abs": 4.584972675030048e-09, | |
| "rel_l2": 2.0996074718879118e-08 | |
| }, | |
| "times": { | |
| "max_abs": 5.960464477539063e-08, | |
| "mean_abs": 4.76837158203125e-09, | |
| "rel_l2": 2.0855193972367177e-08 | |
| }, | |
| "velocity_fp16_vs_fp32ref": { | |
| "max_abs": 0.013125389814376831, | |
| "mean_abs": 0.0010297759424406192, | |
| "rel_l2": 0.0012895214320252043 | |
| }, | |
| "velocity_fp16_vs_fp16ref": { | |
| "max_abs": 0.015636205673217773, | |
| "mean_abs": 0.001322876022487012, | |
| "rel_l2": 0.0016291599890619199 | |
| }, | |
| "ref_fp16_vs_ref_fp32": { | |
| "max_abs": 0.016333699226379395, | |
| "mean_abs": 0.0011320557014777956, | |
| "rel_l2": 0.0013910033166357426 | |
| }, | |
| "velocity_fp32": { | |
| "max_abs": 2.0503997802734375e-05, | |
| "mean_abs": 1.8179455511446113e-06, | |
| "rel_l2": 2.220910126120128e-06 | |
| }, | |
| "decoded_voxels": { | |
| "ours": 1442341, | |
| "reference": 1442339, | |
| "common": 1442333 | |
| }, | |
| "decoded_sdf_sign": { | |
| "max_abs": 28.49630832672119, | |
| "mean_abs": 0.000985199632137544, | |
| "rel_l2": 0.0035812783661513725 | |
| }, | |
| "decoded_feat": { | |
| "max_abs": 0.3575434908270836, | |
| "mean_abs": 2.4584798492266753e-05, | |
| "rel_l2": 0.0023846487551618238 | |
| }, | |
| "mesh_faces_identical": true, | |
| "mesh_vertices": { | |
| "max_abs": 2.86102294921875e-06, | |
| "mean_abs": 2.4002885922643986e-07, | |
| "rel_l2": 5.207820769835567e-08 | |
| }, | |
| "filled_faces_identical": true | |
| }, | |
| "sphere192_report": { | |
| "conditioning": { | |
| "max_abs": 5.714595317840576e-06, | |
| "mean_abs": 2.006490236685449e-07, | |
| "rel_l2": 8.875883576616418e-06 | |
| }, | |
| "sigmas": { | |
| "max_abs": 0.0, | |
| "mean_abs": 0.0, | |
| "rel_l2": 0.0 | |
| }, | |
| "times": { | |
| "max_abs": 0.0, | |
| "mean_abs": 0.0, | |
| "rel_l2": 0.0 | |
| }, | |
| "velocity_fp16_vs_fp32ref": { | |
| "max_abs": 0.032924532890319824, | |
| "mean_abs": 0.0010041114637135014, | |
| "rel_l2": 0.001261798954717289 | |
| }, | |
| "velocity_fp32": { | |
| "max_abs": 5.2988529205322266e-05, | |
| "mean_abs": 2.1241101601153427e-06, | |
| "rel_l2": 2.638211684786904e-06 | |
| } | |
| } | |
| } | |
| }, | |
| "benchmark": { | |
| "hardware": "Apple M4 Pro, 64 GiB unified memory", | |
| "python": "3.12.11", | |
| "input": "Hi3D assets/example.png (436x640 RGBA), coarse mesh from mlx-community/sam-3d-objects-bf16 (seed 0)", | |
| "resolution": 1024, | |
| "steps": 25, | |
| "seed": 0, | |
| "timings": { | |
| "coarse_mesh": 77.6, | |
| "load": 1.5, | |
| "conditioning": 0.58, | |
| "pass1_sampling": 189.08, | |
| "pass1_decoding": 8.86, | |
| "pass2_sampling": 1517.58, | |
| "pass2_decoding": 44.96, | |
| "postprocess": 8.67 | |
| }, | |
| "peak": 30.05, | |
| "vertices": 495425, | |
| "faces": 1000000 | |
| } | |
| } | |