Instructions to use KBlueLeaf/MOSS-Image-FP16-fixed with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use KBlueLeaf/MOSS-Image-FP16-fixed with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("KBlueLeaf/MOSS-Image-FP16-fixed", 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
Commit History
3 days ago
-
Report the 1024^2 plain-fp16 search: no CFG, seed or prompt breaks it
Verified signatureba63601 -
Report LPIPS, SSIM and PSNR for every column; show the fix changes range, not precision
Verified signaturecd13c41 -
Compare fp16 without the fix, fp16 with the fix, bf16 and fp32 in every panel
Verified signature49ef51b -
Add the fp32 2560x2560 image; mark bf16 as re-running with bf16 compute
Verified signaturef3e6fdc -
Keep the card consistent with the uploaded images
Verified signature9377be4 -
Describe where plain fp16 breaks
Verified signaturef5150b8 -
Show plain fp16 going NaN at 2560x2560
Verified signatureae1ea08 -
Charts on a white background
Verified signature24c37e4 -
Count the fold's subnormal and flushed weights over the scaled elements only
Verified signaturea880990 -
Add the fp16-fixed transformers: diffusers (base, sft) and ComfyUI single files
Verified signaturee86a13e -
Add the model card, images and reports
Verified signature9874f33 -
initial commit
Verified signaturee09fc71