Aurora (FLUX-based)

Started as an unchanged copy of Freepik/flux.1-lite-8B (kept under base_original/transformer), then QLoRA fine-tuned and merged into full standalone weights (transformer/). Includes a "Thinker" generator: an LLM expands your idea like a film director, several candidates are painted, and CLIP picks the one that best matches your original idea.

Status: training finished and merged

Repository layout

• model_index.json, transformer/, text_encoder/, text_encoder_2/, tokenizer/, tokenizer_2/, vae/, scheduler/ - the complete model (load with FluxPipeline.from_pretrained)
• base_original/transformer/ - the untouched original transformer weights
• training/ - checkpoints and the merge record
• samples/ - test images
• aurora_generate.py - the Thinker generator

Use

!pip install -q -U diffusers transformers accelerate bitsandbytes sentencepiece peft
from huggingface_hub import hf_hub_download
exec(open(hf_hub_download("AyaanAhmed123/Aurora", "aurora_generate.py")).read())
aurora = Aurora(repo="AyaanAhmed123/Aurora")
image, info = aurora("a lone astronaut walking through a neon rain-soaked city at night")
image.save("out.png")