Instructions to use wfen/Cosmos3-Nano-FP8-Blockwise with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use wfen/Cosmos3-Nano-FP8-Blockwise with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("wfen/Cosmos3-Nano-FP8-Blockwise", 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
Commit History
-
Update README.md
Verified signature379eee9 -
Update README.md
Verified signature8dac185 -
add webui demo info
dfcad08
-
Fix LFS tracking for small text files
Blanket *.json/*.py/*.txt/*.csv/*.jinja rules forced every small config/tokenizer/script file into LFS regardless of size. Removed those rules and renormalized; text_tokenizer/tokenizer.json (11.4 MB) keeps its explicit size-justified LFS override. Large weight files are unaffected (verified: none appear in this commit's diff at all).9bf5d6a -
Remove stale top-level weight index; restore orphaned compliance docs
model.safetensors.index.json referenced 7 non-existent transformer/diffusion_pytorch_model-0000N-of-00007 shards; the real weight is one consolidated transformer/diffusion_pytorch_model.safetensors. BIAS.md/EXPLAINABILITY.md/PRIVACY.md/SAFETY.md were checking out as raw LFS-pointer text (no .gitattributes rule has ever matched .md, so neither a plain clone nor git-lfs's own tooling re-smudges them); restored their real content from the underlying LFS objects.6dfe54e
-
lm_head back to bf16
4e181f9
-
Create README.md
Verified signature5bbfc64 -
Delete README.md
Verified signature574d4e6 -
Update .gitattributes
Verified signature2776483 -
Upload README.md
Verified signatureRe-upload readmece7c09d -
delete readme.md
Verified signatureREADME.md is not supposed to be lfs tracked.6f99aeb -
track mp4 and png files
acb88a0 -
initialize Blockwise FP8 quantization model
739a1f6 -
initial commit
Verified signature2cfbbce