je-suis-tm/melanie_laurent_lora_flux

Melanie Laurent Lora Flux

Prompt

Training With QLoRA: Melanie Laurent, radiating sophistication, sits before a bookshelf in a warmly lit, intimate setting. Dressed in a beige coat with prominent buttons and a matching hat, she holds an open book in her lap, her eyes meeting the viewer's gaze. Her long blonde hair frames her face, contributing to an air of elegance. The composition focuses primarily on the woman, who dominates the frame, while a partially visible bookshelf adds depth to the scene. A single light source from the right creates a soft, cozy atmosphere, enhancing the vintage feel of her attire. The muted color palette and gentle lighting contribute to a sense of refinement and poise, evoking a feeling of exclusivity and quiet grace. The overall impression is one of timeless beauty and understated luxury.

Prompt

Training Without QLoRA: Melanie Laurent, radiating sophistication, sits before a bookshelf in a warmly lit, intimate setting. Dressed in a beige coat with prominent buttons and a matching hat, she holds an open book in her lap, her eyes meeting the viewer's gaze. Her long blonde hair frames her face, contributing to an air of elegance. The composition focuses primarily on the woman, who dominates the frame, while a partially visible bookshelf adds depth to the scene. A single light source from the right creates a soft, cozy atmosphere, enhancing the vintage feel of her attire. The muted color palette and gentle lighting contribute to a sense of refinement and poise, evoking a feeling of exclusivity and quiet grace. The overall impression is one of timeless beauty and understated luxury.

Prompt

Testing With QLoRA: Melanie Laurent with a glowing aura, sphinx with ancient runes, islands floating effortlessly, towering pyramids emitting a soft glow, dusk, magical realism, ((highly detailed) ), egyptian clothes

Prompt

Testing Without QLoRA: Melanie Laurent with a glowing aura, sphinx with ancient runes, islands floating effortlessly, towering pyramids emitting a soft glow, dusk, magical realism, ((highly detailed) ), egyptian clothes

All files are also archived in https://github.com/je-suis-tm/huggingface-archive in case this gets censored.

This a non-quantized version of https://huggingface.co/je-suis-tm/melanie_laurent_lora_flux_nf4. Both are trained on the same dataset. The training is based on https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/test_dreambooth_lora_flux.py. The training took 3 hours on A100 80GB with max VRAM consumption at 35GB. The inference consumes 36GB of VRAM.

Train

    export MODEL_NAME="black-forest-labs/FLUX.1-dev"
    export INSTANCE_DIR="/pvol/melanie_laurent"
    export OUTPUT_DIR="/pvol/melanie_laurent_lora_flux"
    accelerate config default
    accelerate launch train_dreambooth_lora_flux1.py \
      --pretrained_model_name_or_path=$MODEL_NAME  \
      --mixed_precision="bf16" \
      --dataset_name=$INSTANCE_DIR \
      --output_dir=$OUTPUT_DIR \
      --gradient_checkpointing \
      --instance_prompt="Melanie Laurent" \
      --caption_column="text" \
      --resolution=1024 \
      --train_batch_size=1 \
      --guidance_scale=1 \
      --use_8bit_adam \
      --checkpointing_steps=100 \
      --gradient_accumulation_steps=4 \
      --optimizer="adamW" \
      --learning_rate=1e-4 \
      --lr_scheduler="constant" \
      --lr_warmup_steps=100 \
      --max_train_steps=1500 \
      --rank=4 \
      --seed="0" 

Usage

import torch
from diffusers import FluxPipeline

device = "cuda:0"

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

pipe.load_lora_weights("je-suis-tm/melanie_laurent_lora_flux",
                       weight_name='pytorch_lora_weights.safetensors')


prompt = "Melanie Laurent with a glowing aura, sphinx with ancient runes, islands floating effortlessly, towering pyramids emitting a soft glow, dusk, magical realism, ((highly detailed) ), egyptian clothes"

image = pipe(
  prompt=prompt,
  generator=torch.Generator(device=device).manual_seed(42),
  num_inference_steps=50, # 28 is a good trade-off
  guidance_scale=4,
  height=1024, 
  width=1024,
).images[0]

image.save("melanie_laurent.png")

Trigger words

You should use Melanie Laurent to trigger the image generation.

Download model

Download them in the Files & versions tab.