import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("AcademiaSD/MiniMax-H3-NF4", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]MiniMax-H3 — NF4 for LoRA training
NF4 (4-bit) version of MiniMax-H3 (transformer, Qwen3-VL-32B text encoder and VAEs, 41 GB instead of 498 GB), used by AcademiaSD LoRAlab Trainer Studio to train MiniMax-H3 LoRAs and RefMods on consumer GPUs. The trainer downloads it automatically.
These weights are a derivative of MiniMax-H3 and are distributed under the same MiniMax-H3 Community License Agreement. Read it before using the model or the LoRAs trained with it.
- Downloads last month
- 481
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for AcademiaSD/MiniMax-H3-NF4
Base model
MiniMaxAI/MiniMax-H3