README.md

---
base_model: black-forest-labs/flux-3-image-klein-base
library_name: diffusers
license: other
tags:
- flux3
- lora
- diffusers
- text-to-image
- tarot
instance_prompt: a trtcrd, tarot style
widget:
- text: a trtcrd of a person giving a talk on a stage under a glowing emblem, "the speaker" arcana, tarot style
  output:
    url: images/speaker.jpg
- text: a trtcrd of a man carving meat from a tall vertical rotisserie that reads "Döner", "the carver" arcana, tarot style
  output:
    url: images/carver.jpg
- text: a trtcrd of a dog eating a big bowl of spaghetti at a dinner table, "the glutton" arcana, tarot style
  output:
    url: images/glutton.jpg
- text: a trtcrd of an astronaut planting a flag on a tiny planet among stars, "the explorer" arcana, tarot style
  output:
    url: images/explorer.jpg
---

# FLUX.3 Tarot LoRA

A Rider-Waite tarot-style DreamBooth LoRA for the **FLUX.3 image (klein) model**, trained with the diffusers DreamBooth FLUX.3 trainer. Trigger the style with **`trtcrd`**, e.g. `"a trtcrd of ..., tarot style"`.

<Gallery />

## Inference

```python
import torch
from diffusers.modular_pipelines import Flux3AutoBlocks

pipe = Flux3AutoBlocks().init_pipeline("black-forest-labs/flux-3-image-klein-base")
pipe.load_components(dtype=torch.bfloat16)
pipe.to("cuda")
pipe.load_lora_weights("diffusers-modular/flux3-tarot-lora", weight_name="pytorch_lora_weights.safetensors")

image = pipe(
    prompt="a trtcrd of a dog eating a big bowl of spaghetti, tarot style",
    height=1344, width=800, num_inference_steps=28, negative_prompt="",
    max_sequence_length=512, text_encoder_out_layers=(4, 8, 12, 16, 20, 24, 28, 32),
    output="images",
)[0]
image.save("tarot.png")
```

## Trigger

Use `trtcrd` in the prompt: `"a trtcrd of {your subject}, tarot style"`.

## Training

- Base model: `black-forest-labs/flux-3-image-klein-base`
- Dataset: [multimodalart/1920-raider-waite-tarot-public-domain](https://huggingface.co/datasets/multimodalart/1920-raider-waite-tarot-public-domain)
- 1500 steps · seed 0 · resolution 1024 · aspect-ratio buckets
- Default weights (`pytorch_lora_weights.safetensors`): AdamW (8-bit), rank 16, lr 1e-4

### Sweep variants

This repo also includes the configs from a small sweep (one setting changed vs the default). Load one with `weight_name="<folder>/pytorch_lora_weights.safetensors"`:

| variant | change | `weight_name` |
|---|---|---|
| default | AdamW, rank 16 | `pytorch_lora_weights.safetensors` |
| rank 8 | rank → 8 | `rank8/pytorch_lora_weights.safetensors` |
| rank 32 | rank → 32 | `rank32/pytorch_lora_weights.safetensors` |
| more layers | + modulation layers | `more-layers/pytorch_lora_weights.safetensors` |
| prodigy | Prodigy optimizer, lr 1 | `prodigy/pytorch_lora_weights.safetensors` |