Text-to-Image
MLX
MiniMax H3
Safetensors
English
apple-silicon
macos
image-to-image
image-editing
image-generation
Instructions to use devin-lai/MiniMax-H3-Image-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use devin-lai/MiniMax-H3-Image-mlx with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download devin-lai/MiniMax-H3-Image-mlx --local-dir MiniMax-H3-Image-mlx
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Download optimization/round-03/contract.json from devin-lai/MiniMax-H3-Image-mlx: direct link, hf CLI and curl.
- Browser
- Download file 1.33 kB
-
https://huggingface.co/devin-lai/MiniMax-H3-Image-mlx/resolve/main/optimization/round-03/contract.json
- Command line
-
hf download hf://devin-lai/MiniMax-H3-Image-mlx/optimization/round-03/contract.json
-
curl -L -o contract.json https://huggingface.co/devin-lai/MiniMax-H3-Image-mlx/resolve/main/optimization/round-03/contract.json
1.33 kB
| { | |
| "schema_version": 1, | |
| "campaign_id": "h3-m5pro-round-03", | |
| "objective": "latency", | |
| "quality_cases": [ | |
| "product", | |
| "portrait", | |
| "edit", | |
| "holdout-detail" | |
| ], | |
| "timing_cases": [ | |
| "product", | |
| "portrait" | |
| ], | |
| "quality_gates": { | |
| "max_abs_error": { | |
| "direction": "max", | |
| "limit": 0 | |
| }, | |
| "nonfinite_values": { | |
| "direction": "max", | |
| "limit": 0 | |
| } | |
| }, | |
| "workload_manifest_sha256": "5dfe34824190abd241185fd37214e3de5be425d82f0a6f556d0d23f0dec68042", | |
| "reference_sha256": "201be3382041aec890cbd766c2f88a43799f7e2d33807e9a1504f7e29e422065", | |
| "evaluator_sha256": "4db0c5e8606ffc011bf85dbadabc482468a2e49cd5a1757a774620d493eba7ee", | |
| "min_repeats": 3, | |
| "min_latency_reduction_pct": 2, | |
| "min_memory_reduction_pct": 3, | |
| "max_latency_regression_pct": 0, | |
| "max_memory_regression_pct": 1, | |
| "peak_memory_limit_bytes": 53687091200, | |
| "search_budget": "120 minutes from 08:02 UTC, at most 72 MPP GEMM tiles, 6 attention layouts and 12 unpack settings, three full-image finalists, one GPU job at a time, <=5 GiB additional disk. Exact track first. Early reject on no gain. Parent process physical footprint sampled at 100ms; allocator peak reported separately. Warm paired product/portrait repeats, alternate AB/BA; first-load separately. No background user applications changed." | |
| } | |