FLUX Image Generator Setup | Generated by AI
All set! Here’s what’s in the flux/ directory:
| File | Purpose |
|---|---|
run_flux_schnell.py |
Main script — download & generate images |
requirements.txt |
Python dependencies (diffusers, torch, etc.) |
README.md |
Full documentation |
Quick start
cd /mnt/data/zz/flux
# Install deps
python3 -m pip install --break-system-packages -r requirements.txt
# Run with your HF token
python3 run_flux_schnell.py --token hf_YOUR_TOKEN_HERE
What the script does
- Downloads the 12B‑parameter FLUX.1-schnell model from Hugging Face
- Generates images from your text prompts (default: cat with sign, cyberpunk street, fantasy dragon)
- Saves PNGs to
./outputs/
Important
⚠️ The model is gated — you must:
- Go to https://huggingface.co/black-forest-labs/FLUX.1-schnell
- Click “Agree and access repository”
- Create a token at https://huggingface.co/settings/tokens
Then pass it via --token hf_xxxx or export HUGGING_FACE_HUB_TOKEN=hf_xxxx.
Key args
| Arg | Default | Description |
|---|---|---|
prompts... |
3 defaults | Text prompts |
--token |
env var | Hugging Face token |
--steps |
4 | Inference steps (1–4) |
--seed |
0 | Base seed (auto‑incremented per prompt) |
The script uses bfloat16 + CPU offloading to fit the 12B model in limited VRAM. Schnell is distilled — it produces great results in just 1–4 steps with guidance_scale=0.0.
Generated by AI. Curating and sharing still takes effort.
If you find it useful, feel free to
donate.
WeChat: @lzwjava
·
X: @lzwjava ·
Say hi 👋