lora-monet-sd1.5

lora-monet-sd1.5

LoRA adapter for Monet-style image generation on top of Stable Diffusion 1.5.

Model summary

• Base model: runwayml/stable-diffusion-v1-5
• Trigger word: msl monet
• Adapter file: pytorch_lora_weights.safetensors
• Intended style: soft brush strokes, impressionist color palette, diary-like painterly scenes

Intended use

This adapter is intended for stylized image generation, visual diary experiments, and Monet-inspired scenery or illustration prompts.

It works best when used as a style adapter on top of SD1.5 rather than as a general-purpose object model.

Related project

• Companion app repo: https://github.com/J-Y00N/Multimodal-Picture-Diary
• Live demo page: https://multimodal-picture-diary-25vtwte77q6nxwtbdhhpp9.streamlit.app/

This model repo is maintained by the same author as the companion app, but published separately so the LoRA release and the application code can be versioned independently.

Preserved training settings

The following settings are preserved from the surviving training config:

• resolution: 1024x1024
• network_dim: 8
• network_alpha: 1
• train_batch_size: 16
• text_encoder_lr: 5e-05
• unet_lr: 0.0001
• optimizer: AdamW
• epochs: 100
• xformers enabled

Preserved training artifacts indicate the adapter was originally trained with a kohya_ss-based LoRA workflow.

See metadata/monet_last_20231129-043745.json for the preserved config snapshot.

Dataset reference

• Public dataset reference: Kaggle - GAN Getting Started / Monet-style dataset

This link is included as the public dataset reference most closely associated with the preserved Monet-style training experiment.

Known unknowns

• Exact training seed was not preserved.
• Exact dataset snapshot and caption corpus are not fully preserved.
• This repository is an inference-oriented adapter release, not a full archival dump of the original training environment.

Diffusers usage

import torch
from diffusers import StableDiffusionPipeline

pipe = StableDiffusionPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    torch_dtype=torch.float16,
).to("cuda")

pipe.load_lora_weights(
    "J-YOON/lora-monet-sd1.5",
    weight_name="pytorch_lora_weights.safetensors",
)

prompt = "msl monet, a rainy bridge at sunset, diary illustration, soft brush strokes"
image = pipe(prompt, num_inference_steps=30, guidance_scale=7.0).images[0]
image.save("monet_example.png")

Limitations

• The adapter is style-specialized and may bias outputs toward painterly scenery or impressionist compositions.
• General object fidelity can drop for prompts far from the original training domain.
• For broader prompt coverage, reduce adapter strength or fall back to the base model when needed.

Gallery

<table> <tr> <td align="center"><strong>Generated: Dog</strong></td> <td align="center"><strong>Generated: Diary img2img</strong></td> </tr> <tr> <td><img src="./images/example_dog_monet.png" alt="Monet dog example" width="260" /></td> <td><img src="./images/example_rainy_walk_monet.png" alt="Diary app img2img example" width="260" /></td> </tr> <tr> <td align="center"><strong>Reference: Dog</strong></td> <td align="center"><strong>Reference: Rainy walk</strong></td> </tr> <tr> <td><img src="./images/reference_dog_photo.jpg" alt="Reference dog photo" width="260" /></td> <td><img src="./images/reference_rainy_walk_photo.jpg" alt="Reference rainy walk photo" width="260" /></td> </tr> </table>