Instructions to use TDSXn9hai/exflux with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use TDSXn9hai/exflux with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("TDSXn9hai/exflux", 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
import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("TDSXn9hai/exflux", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]EC-FLUX Stage 2 v2
Structured-pruned FLUX.2 Klein 4B Base DiT checkpoints from the EC-FLUX GitHub repository.
Only three selected checkpoints are published:
| Directory | DiT parameters | Parameter ratio | Removed Single Blocks |
|---|---|---|---|
full |
2.891B | 0.746 | 6, 8, 10, 13 |
fixed_ratio |
3.227B | 0.833 | 6, 8, 10, 13 |
no_block |
3.340B | 0.862 | none |
Each directory contains:
model.safetensors: pruned DiT weights.prune_spec.json: original unit indices and reconstruction metadata.history.json: APRA rounds, gates, recovery steps, and final metrics.run_config.json: experiment configuration.eval_summary.jsonandeff_summary.json: D_val evaluation and 512px efficiency measurements.
Use the original FLUX.2 Klein 4B pipeline for the text encoder, tokenizer, VAE,
and scheduler, replacing its transformer with one of these pruned DiT
checkpoints. The repository supports loading via ecflux.flux_utils.load_transformer.
The checkpoints inherit the FLUX.2 Klein non-commercial license and acceptable use policy. See the upstream model: https://huggingface.co/black-forest-labs/FLUX.2-klein-base-4B.
Source results
- Downloads last month
- -