Image-Text-to-Video
Diffusers
Safetensors
orbitquant
comfyui
w4
w4a4
native-w4a4-transformer-runtime
text-to-video
audio-video-generation
8-bit precision
Instructions to use WaveCut/MiniMax-H3-OrbitQuant-W4A4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use WaveCut/MiniMax-H3-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/MiniMax-H3-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
| { | |
| "status": "pass", | |
| "task": "ref2va", | |
| "component": "transformer_ref", | |
| "runtime": "packed_w4_dequant_bf16_without_persistent_dequant_cache", | |
| "width": 608, | |
| "height": 480, | |
| "num_frames": 124, | |
| "num_inference_steps": 50, | |
| "planned_model_evaluations": 49, | |
| "completed_model_evaluations": 1, | |
| "first_forward_seconds": 212.94, | |
| "observed_nvidia_smi_peak_mib": 38577, | |
| "checkpoint": "step-001-of-049.pt", | |
| "checkpoint_tensor_count": 108, | |
| "all_checkpoint_tensors_finite": true, | |
| "full_video_status": "not_run", | |
| "full_video_reason": "At the measured first-forward rate, a 49-forward Ref2VA smoke would take approximately 2 hours 54 minutes before FP32 VAE decode. This release does not claim an unobserved full Ref2VA result." | |
| } | |