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 reports/optimization-r3/compact-deployment/control.receipt.json from devin-lai/MiniMax-H3-Image-mlx: direct link, hf CLI and curl.
- Browser
- Download file 5.47 kB
-
https://huggingface.co/devin-lai/MiniMax-H3-Image-mlx/resolve/main/reports/optimization-r3/compact-deployment/control.receipt.json
- Command line
-
hf download hf://devin-lai/MiniMax-H3-Image-mlx/reports/optimization-r3/compact-deployment/control.receipt.json
-
curl -L -o control.receipt.json https://huggingface.co/devin-lai/MiniMax-H3-Image-mlx/resolve/main/reports/optimization-r3/compact-deployment/control.receipt.json
5.47 kB
| { | |
| "schema_version": 1, | |
| "id": "r3-deployment-control", | |
| "campaign_id": "h3-m5pro-round-03-compact", | |
| "contract_sha256": "e95ad9cbe3e6251bc050377cb2e42e681b99d308a30d6a54decef0d8ce2d7b31", | |
| "complete": true, | |
| "change_class": "control", | |
| "hypothesis": "Native reduction matching QK/RoPE fusion and shape-qualified Metal 4 FC2 tiling reduce full generation latency without changing a denoising tensor or pixel.", | |
| "context": { | |
| "hardware": "Apple M5 Pro, 16 GPU cores, 64 GiB unified memory", | |
| "os": "macOS 27.0.1 arm64", | |
| "software": "MLX/MLX Metal 0.32.4; Python 3.12.14; strict precision environment", | |
| "environment": "AC Automatic power; MLX_ENABLE_TF32=0, MLX_CONV_WINOGRAD=0; user background apps unchanged; no thermal warning recorded", | |
| "timing_boundary": "Pipeline.generate total_s: warm text cache, original CPU RNG, six heads, decode, trace and PNG save. Excludes interpreter startup and external quality scoring.", | |
| "memory_metric": "Parent macOS physical footprint sampled at 100ms, current-interval maximum; includes Metal, excludes compiler XPC. MLX allocator separately; do not sum. Both compiled variants may be resident.", | |
| "cache_state": "Resident shared native weights and VAE; shape warmup for each variant. Exact conditioning caches warmed outside timed population.", | |
| "execution_policy": "Single GPU benchmark process; per-case repeats alternate control/candidate then candidate/control then control/candidate. Warmup and first-load excluded from timed population." | |
| }, | |
| "variant": { | |
| "id": "control", | |
| "code_sha256": "5be81c114b17d671881c9d0323d8c305e10fa450ad228a36efd4a264892e6d3d", | |
| "model_sha256": "3274a3e5fe0d2ea0d63f66b020beb193276a0a9d8de7ea466b312fbae9a2746f", | |
| "config_sha256": "da98e53b10f54d4398305321b423e14df104bb4f4c08c5c363731ced0113d931", | |
| "command": [ | |
| ".venv/bin/python", | |
| "-m", | |
| "experiments.r3.images", | |
| "--deployment", | |
| "--variants", | |
| "qk-fc2-native-norm", | |
| "--workload", | |
| "optimization/round-03-compact/workload.json", | |
| "--model", | |
| "/Users/devin/Desktop/minimax-h3-image-opt/models/MiniMax-H3-Image", | |
| "--text-encoder", | |
| "/Users/devin/Desktop/minimax-h3-image-opt/models/MiniMax-H3/text_encoder", | |
| "--output", | |
| "outputs/optimization-r3-compact" | |
| ] | |
| }, | |
| "quality": { | |
| "variant_sha256": "751869812828b19f54f3907c6b01cdec4afd35682127b86c01d1459d78401960", | |
| "reference_sha256": "5fccbc8664a6efb3f9ff7e14b764c1aae876785fc28fd492dbb6c24b1b773a8b", | |
| "evaluator_sha256": "4db0c5e8606ffc011bf85dbadabc482468a2e49cd5a1757a774620d493eba7ee", | |
| "outputs": { | |
| "product": "eaed6c290ea3f7843ed4a789614cf3ad384d5acf3d0e0cc40a569fdcb850390f", | |
| "portrait": "4226310bd1d79a0d0d4c0b572e2e7729541f5f171260cfeefd5e73b0a4196a47", | |
| "edit": "b4b5875fea6451b9c7c24efc7903945a454ae7885ed3016bfc361b4799b2508e", | |
| "holdout-detail": "561b18aa708d61a5accbfd12f2280e8f7e303505324c43471ba06d60fe0351d5" | |
| }, | |
| "metrics": { | |
| "max_abs_error": 0, | |
| "nonfinite_values": 0 | |
| } | |
| }, | |
| "runs": [ | |
| { | |
| "case": "product", | |
| "repeat": 0, | |
| "seconds": 66.73659237497486, | |
| "peak_memory_bytes": 34106662192, | |
| "output_sha256": "eaed6c290ea3f7843ed4a789614cf3ad384d5acf3d0e0cc40a569fdcb850390f" | |
| }, | |
| { | |
| "case": "product", | |
| "repeat": 1, | |
| "seconds": 67.69735000003129, | |
| "peak_memory_bytes": 34232442256, | |
| "output_sha256": "eaed6c290ea3f7843ed4a789614cf3ad384d5acf3d0e0cc40a569fdcb850390f" | |
| }, | |
| { | |
| "case": "product", | |
| "repeat": 2, | |
| "seconds": 66.51806045905687, | |
| "peak_memory_bytes": 34269748624, | |
| "output_sha256": "eaed6c290ea3f7843ed4a789614cf3ad384d5acf3d0e0cc40a569fdcb850390f" | |
| }, | |
| { | |
| "case": "portrait", | |
| "repeat": 0, | |
| "seconds": 66.77446975000203, | |
| "peak_memory_bytes": 34397462120, | |
| "output_sha256": "4226310bd1d79a0d0d4c0b572e2e7729541f5f171260cfeefd5e73b0a4196a47" | |
| }, | |
| { | |
| "case": "portrait", | |
| "repeat": 1, | |
| "seconds": 66.11278283409774, | |
| "peak_memory_bytes": 34380471936, | |
| "output_sha256": "4226310bd1d79a0d0d4c0b572e2e7729541f5f171260cfeefd5e73b0a4196a47" | |
| }, | |
| { | |
| "case": "portrait", | |
| "repeat": 2, | |
| "seconds": 65.95004508295096, | |
| "peak_memory_bytes": 34352488040, | |
| "output_sha256": "4226310bd1d79a0d0d4c0b572e2e7729541f5f171260cfeefd5e73b0a4196a47" | |
| } | |
| ], | |
| "evidence": [ | |
| { | |
| "path": "outputs/optimization-r3-compact/runs.jsonl", | |
| "sha256": "fdd30c091a2f2eaf01e5b2facc32d146796152d560701c472b37c9b226dfe7e3" | |
| }, | |
| { | |
| "path": "outputs/optimization-r3-compact/source-manifest.json", | |
| "sha256": "5be81c114b17d671881c9d0323d8c305e10fa450ad228a36efd4a264892e6d3d" | |
| }, | |
| { | |
| "path": "reports/optimization-r3/compact-deployment/quality-audit.json", | |
| "sha256": "57ae4aef9579058528381de211636fe9ff02ad9b2f793f65b7b217e91c6e7f16" | |
| }, | |
| { | |
| "path": "reports/optimization-r3/compact-deployment/summary.json", | |
| "sha256": "020ad62fe1e4ad92331b2e86fb63caa18462adad65e12a86ebadf5b751fa607d" | |
| } | |
| ], | |
| "extensions": { | |
| "canonical_output": "SHA256 decoded RGB uint8 bytes in row-major order", | |
| "model_identity": "model_sha256 is the pinned model/source provenance manifest hash; no new full checkpoint rehash", | |
| "reference_scope": "Incremental identity to MLX BF16 recipe, not CUDA or unpublished FP32 training weights", | |
| "trace_identical": true | |
| } | |
| } | |