Instructions to use WaveCut/FLUX.2-klein-9B-OrbitQuant-W4A4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use WaveCut/FLUX.2-klein-9B-OrbitQuant-W4A4 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("WaveCut/FLUX.2-klein-9B-OrbitQuant-W4A4", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
| { | |
| "schema_version": 1, | |
| "model": "FLUX.2 Klein 9B", | |
| "note": "single-host three-variant comparison: one GPU, one session, separate processes per variant, pinned revisions", | |
| "hardware": { | |
| "gpu": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition", | |
| "torch": "2.9.1+cu128", | |
| "cuda": "12.8" | |
| }, | |
| "protocol": { | |
| "width": 1024, | |
| "height": 1024, | |
| "steps": 4, | |
| "guidance": 1.0, | |
| "seed": 0, | |
| "prompt_count": 10, | |
| "dtype": "bfloat16", | |
| "cpu_offload": false, | |
| "allocator": "expandable_segments" | |
| }, | |
| "orbitquant_package_version": "0.5.0", | |
| "variants": [ | |
| { | |
| "variant": "bf16", | |
| "repo": "black-forest-labs/FLUX.2-klein-9B", | |
| "revision": "92196c8e11f7b6cf2b7493e037d8c5345c559216", | |
| "gpu": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition", | |
| "torch": "2.9.1+cu128", | |
| "load_seconds": 3.061781742144376, | |
| "load_nvml_peak_bytes": 35979329536, | |
| "cold_seconds": 1.5380592348519713, | |
| "hot_mean_seconds": 1.2014360929839314, | |
| "hot_median_seconds": 1.201876456500031, | |
| "hot_p95_seconds": 1.207286735996604, | |
| "max_cuda_allocated_peak_bytes": 37284590080, | |
| "max_cuda_reserved_peak_bytes": 37708890112, | |
| "max_nvml_peak_bytes": 39081017344 | |
| }, | |
| { | |
| "variant": "sdnq", | |
| "repo": "WaveCut/FLUX.2-klein-9B-SDNQ-uint4-static", | |
| "revision": "ed71b3f19ce640e88b66a2a743aabb8a613adeac", | |
| "gpu": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition", | |
| "torch": "2.9.1+cu128", | |
| "load_seconds": 3.0232842238619924, | |
| "load_nvml_peak_bytes": 13451722752, | |
| "cold_seconds": 5.8073041120078415, | |
| "hot_mean_seconds": 1.2734668201068415, | |
| "hot_median_seconds": 1.272957639535889, | |
| "hot_p95_seconds": 1.2804428350646049, | |
| "max_cuda_allocated_peak_bytes": 14759080448, | |
| "max_cuda_reserved_peak_bytes": 15179186176, | |
| "max_nvml_peak_bytes": 16553410560 | |
| }, | |
| { | |
| "variant": "orbitquant", | |
| "repo": "WaveCut/FLUX.2-klein-9B-OrbitQuant-W4A4", | |
| "revision": "ee3a38f7767ae199818d746c840be0f1837887bf", | |
| "gpu": "NVIDIA RTX PRO 6000 Blackwell Workstation Edition", | |
| "torch": "2.9.1+cu128", | |
| "load_seconds": 1.946834340924397, | |
| "load_nvml_peak_bytes": 12109545472, | |
| "cold_seconds": 4.956948169972748, | |
| "hot_mean_seconds": 1.163340676180087, | |
| "hot_median_seconds": 1.163962614024058, | |
| "hot_p95_seconds": 1.1686886479146779, | |
| "max_cuda_allocated_peak_bytes": 13429025280, | |
| "max_cuda_reserved_peak_bytes": 13841203200, | |
| "max_nvml_peak_bytes": 15215427584 | |
| } | |
| ] | |
| } |