---
tags:
- text-to-image
- lora
- diffusers
- template:diffusion-lora
widget:
- text: 'A 3D Portrait cartoon depiction of a woman with long dark brown hair and brown eyes stands in front of a window. She is wearing a gray turtleneck, with her left hand resting on her head. Her right hand is touching her face, adding a touch of warmth to the scene. The background is a stark white wall, with a window on the left side of the image.'
  output:
    url: images/1.webp
- text: 'A 3D cartoon-style portrait of a man with short, wavy black hair and blue eyes, standing against a soft pastel-colored background. He is wearing a casual navy blue hoodie, with his right arm crossed over his chest, and his left hand raised to touch his chin thoughtfully. The background includes a large, arched window with light streaming through, creating a warm, cozy atmosphere'
  output:
    url: images/2.webp
- text: 'A 3D cartoon-style portrait of a young woman with medium-length blonde hair and green eyes, sitting in a cozy cafe. She is wearing a light pink sweater and holding a warm cup of coffee close to her face with both hands, creating a peaceful vibe. The background shows a wooden table and a window with raindrops, giving a cozy, rainy-day feel'
  output:
    url: images/3.webp
base_model: black-forest-labs/FLUX.1-dev
instance_prompt: 3D Portrait
license: creativeml-openrail-m
---
# Castor-3D-Portrait-Flux-LoRA

<Gallery />
**The model is still in the training phase. This is not the final version and may contain artifacts and perform poorly in some cases.**

## Model description 

**prithivMLmods/Castor-3D-Portrait-Flux-LoRAA**

Image Processing Parameters 

| Parameter                 | Value  | Parameter                 | Value  |
|---------------------------|--------|---------------------------|--------|
| LR Scheduler              | constant | Noise Offset              | 0.03   |
| Optimizer                 | AdamW  | Multires Noise Discount   | 0.1    |
| Network Dim               | 64     | Multires Noise Iterations | 10     |
| Network Alpha             | 32     | Repeat & Steps           | 20 & 2.5K |
| Epoch                     | 15     | Save Every N Epochs       | 1      |

    Labeling: florence2-en(natural language & English)
    
    Total Images Used for Training : 19

## Setting Up
```
import torch
from pipelines import DiffusionPipeline

base_model = "black-forest-labs/FLUX.1-dev"
pipe = DiffusionPipeline.from_pretrained(base_model, torch_dtype=torch.bfloat16)

lora_repo = "prithivMLmods/Castor-3D-Portrait-Flux-LoRA"
trigger_word = "3D Portrait"  # Leave trigger_word blank if not used.
pipe.load_lora_weights(lora_repo)

device = torch.device("cuda")
pipe.to(device)
```
## App File Structure

/project-root/

├── .gitattributes         
├── README.md      
├── app.py             
├── pythonproject.py     

## Trigger words 🧨

> [!WARNING]
> **Trigger words:**  You should use `3D Portrait` to trigger the image generation.

## Download model

Weights for this model are available in Safetensors format.

[Download](/prithivMLmods/Castor-3D-Portrait-Flux-LoRA/tree/main) them in the Files & versions tab.