Text-to-Image
Diffusers
Safetensors
MiniMax H3
English
orbitquant
w4a8
4-bit precision
gptq
quantized
image-editing
Instructions to use WaveCut/MiniMax-H3-Image-OrbitQuant-W4A8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use WaveCut/MiniMax-H3-Image-OrbitQuant-W4A8 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-Image-OrbitQuant-W4A8", 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
Download transformer/quantization_config.json from WaveCut/MiniMax-H3-Image-OrbitQuant-W4A8: direct link, hf CLI and curl.
- Browser
- Download file 1.75 kB
-
https://huggingface.co/WaveCut/MiniMax-H3-Image-OrbitQuant-W4A8/resolve/main/transformer/quantization_config.json
- Command line
-
hf download hf://WaveCut/MiniMax-H3-Image-OrbitQuant-W4A8/transformer/quantization_config.json
-
curl -L -o quantization_config.json https://huggingface.co/WaveCut/MiniMax-H3-Image-OrbitQuant-W4A8/resolve/main/transformer/quantization_config.json
1.75 kB
| { | |
| "quant_method": "orbitquant", | |
| "weight_bits": 4, | |
| "activation_bits": 8, | |
| "rotation": "rpbh", | |
| "rotation_seed": 0, | |
| "block_size": "paper", | |
| "rotation_rounds": "auto", | |
| "codebook": "lloyd_max", | |
| "codebook_version": 3, | |
| "codebook_dtype": "float32", | |
| "row_norm_dtype": "bfloat16", | |
| "activation_norm_dtype": "float32", | |
| "activation_eps": 1e-10, | |
| "activation_quantizer": "absmax_int8", | |
| "weight_pack_dtype": "uint8", | |
| "target_policy": "universal", | |
| "target_policy_version": 2, | |
| "adaln_policy": "int4_rtn", | |
| "adaln_group_size": 64, | |
| "adaln_scale": "absmax", | |
| "output_gain_correction": false, | |
| "modules_to_convert": [ | |
| "transformer_blocks.*.attn.to_q", | |
| "transformer_blocks.*.attn.to_k", | |
| "transformer_blocks.*.attn.to_v", | |
| "transformer_blocks.*.attn.to_out.0", | |
| "transformer_blocks.*.ff.net.0.proj", | |
| "transformer_blocks.*.ff.net.2", | |
| "token_refiner.refiner_blocks.*.attn.to_q", | |
| "token_refiner.refiner_blocks.*.attn.to_k", | |
| "token_refiner.refiner_blocks.*.attn.to_v", | |
| "token_refiner.refiner_blocks.*.attn.to_out.0", | |
| "token_refiner.refiner_blocks.*.ff.net.0.proj", | |
| "token_refiner.refiner_blocks.*.ff.net.2" | |
| ], | |
| "modules_to_use_adaln": [], | |
| "modules_to_not_convert": [], | |
| "modules_dtype_dict": {}, | |
| "outlier_channels": {}, | |
| "artifact_format_version": 1, | |
| "runtime_mode": "auto_fused", | |
| "activation_kernel_backend": "auto", | |
| "packed_matmul_block_m": 64, | |
| "packed_matmul_block_n": 64, | |
| "packed_matmul_block_k": 128, | |
| "packed_matmul_num_warps": 4, | |
| "weight_row_tile_size": 256, | |
| "w4a4_int8_weight_cache": false, | |
| "lowbit_boundary_protection": "auto", | |
| "lowbit_protected_blocks": 4, | |
| "lowbit_protected_bits": 4, | |
| "lowbit_interior_protection": "auto", | |
| "fused_layout": null | |
| } | |