Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,20 @@
+---
+license: other
+base_model: black-forest-labs/FLUX.2-klein-4B
+tags: [lora, colorization, flux2]
+---
+# Flux2-3500K — FLUX.2-klein-4B colorize LoRA (runD, 3500 steps)
+
+Grayscale -> color img2img LoRA voor FLUX.2-klein-4B. Beste checkpoint ("runD-3500").
+
+**Recept:** rank 32, lora_alpha 32, lr 1e-4 cosine (warmup 200), weighting_scheme logit_normal (mean -1.0), guidance 1.0 (train=eval), 3500 steps, resolutie 512, img2img (grijs = conditioning). Getraind op recipe_A modern color/gray paren + Spiegel v1-anchor.
+
+**Gebruik:**
+```python
+from diffusers import Flux2KleinPipeline
+import torch
+pipe = Flux2KleinPipeline.from_pretrained('black-forest-labs/FLUX.2-klein-4B', torch_dtype=torch.bfloat16).to('cuda')
+pipe.load_lora_weights('HaaDeej/Flux2-3500K')
+out = pipe(image=gray_square, prompt_embeds=cached_embed, num_inference_steps=4, guidance_scale=1.0).images[0]
+```
+Prompt: "colorize this black and white photograph with natural, realistic colors" (~4 steps, ~3.5s/frame op RTX 4090).