Downloads · 30 days
64
22% of all-time downloads
je-suis-tm/wong_kar_wai_fallen_angels_lora_flux
wong_kar_wai_fallen_angels_lora_flux is a text-to-image model from je-suis-tm. Use it when you need an image from a text prompt. It is set up for diffusers. The card lists the license as mit.
Downloads · 30 days
64
22% of all-time downloads
All-time downloads
297
Public
Repo size
16.2 MB
Likes
2
Public
Click a slice to open those files.
.safetensors9.4 MB · 58%
From the Hugging Face model README
All files are also archived in https://github.com/je-suis-tm/huggingface-archive in case this gets censored.
Wong Kar Wai is one of my fav directors. This LoRA intends to replicate that bold neon color tone, extreme wide angle, surreal neo-noir vibe of the movie Fallen Angels (1995).
The training is based on https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth_lora_flux.py. The training took 2 hours on A100 80GB with max VRAM consumption at 35GB. The inference consumes 36GB of VRAM.
export MODEL_NAME="black-forest-labs/FLUX.1-dev"
export INSTANCE_DIR="/pvol/wong_kar_wai_fallen_angels"
export OUTPUT_DIR="/pvol/wong_kar_wai_fallen_angels_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="WKW style" \
--caption_column="text" \
--resolution=800 \
--center_crop \
--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"
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/wong_kar_wai_fallen_angels_lora_flux",
weight_name='pytorch_lora_weights.safetensors')
prompt = "WKW Style, a tilted view of a woman with long, dark
hair, clad in a blue dress, stands within a subway train. Her hand reaches
up to a metal overhead handle as she raises her chin, seemingly lost in
thought. her eyes are obscured by her bangs. The background suggests an
older subway train model, with indistinct windows and seats visible. The
scene evokes a quiet, peaceful atmosphere, hinting at a solitary journey or
a moment of reflection while waiting for her stop."
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=800,
width=1456, #the movie was in a weird resolution 1056*1912 but flux1 only supports square aspect ratio training
).images[0]
image.save("wong_kar_wai_fallen_angels.png")
You should use WKW style to trigger the image generation.
Download them in the Files & versions tab.