je-suis-tm/sky_bri_lora_flux_nf4

Sky Bri Lora Flux NF4

Prompt

Training With QLoRA: A fashion or lifestyle photograph captures Sky Bri in a sun-drenched outdoor setting. She poses confidently, showcasing a white swimsuit with a plunging neckline and stylish cutouts, complemented by a delicate silver necklace with a pendant. Her long blonde hair cascades around her. The backdrop hints at a tropical or subtropical location, with lush palm trees visible. A black metal railing provides a horizontal element behind the subject. The bright, warm lighting and overall composition suggest a daytime shoot in a pleasant climate. The image's focus is clearly on highlighting the swimsuit and accessories, presenting a stylish and aspirational aesthetic.

Prompt

Training Without QLoRA: A fashion or lifestyle photograph captures Sky Bri in a sun-drenched outdoor setting. She poses confidently, showcasing a white swimsuit with a plunging neckline and stylish cutouts, complemented by a delicate silver necklace with a pendant. Her long blonde hair cascades around her. The backdrop hints at a tropical or subtropical location, with lush palm trees visible. A black metal railing provides a horizontal element behind the subject. The bright, warm lighting and overall composition suggest a daytime shoot in a pleasant climate. The image's focus is clearly on highlighting the swimsuit and accessories, presenting a stylish and aspirational aesthetic.

Prompt

Testing With QLoRA: Sky Bri posing on a luxury yacht, stylish swimwear, ocean waves in the background, glamorous vacation vibe, capturing high-end travel content for social media.

Prompt

Testing Without QLoRA: Sky Bri posing on a luxury yacht, stylish swimwear, ocean waves in the background, glamorous vacation vibe, capturing high-end travel content for social media.

The QLoRA fine-tuning process of sky_bri_lora_flux_nf4 takes inspiration from this post (https://huggingface.co/blog/diffusers-quantization). The training was executed on a local computer with 1000 timesteps and the same parameters as the link mentioned above, which took around 6 hours on 8GB VRAM 4060. The peak VRAM usage was around 7.7GB. To avoid running low on VRAM, both transformers and text_encoder were quantized. One of the issues with Sky Bri is most of her pics are from Instagram which are heavily filtered. This creates a very unrealistic skin. All the images generated here are using the below parameters

• Height: 512
• Width: 512
• Guidance scale: 5
• Num inference steps: 20
• Max sequence length: 512
• Seed: 0

Usage

import torch
from diffusers import FluxPipeline, FluxTransformer2DModel
from transformers import T5EncoderModel

text_encoder_4bit = T5EncoderModel.from_pretrained(
    "hf-internal-testing/flux.1-dev-nf4-pkg", subfolder="text_encoder_2",torch_dtype=torch.float16,)

transformer_4bit = FluxTransformer2DModel.from_pretrained(
        "hf-internal-testing/flux.1-dev-nf4-pkg", subfolder="transformer",torch_dtype=torch.float16,)

pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=torch.float16,
                                    transformer=transformer_4bit,text_encoder_2=text_encoder_4bit)

pipe.load_lora_weights("je-suis-tm/sky_bri_lora_flux_nf4",
                       weight_name='pytorch_lora_weights.safetensors')

prompt="Sky Bri posing on a luxury yacht, stylish swimwear, ocean waves in the background, glamorous vacation vibe, capturing high-end travel content for social media."

image = pipe(
            prompt,
            height=512,
            width=512,
            guidance_scale=5,
            num_inference_steps=20,
            max_sequence_length=512,
            generator=torch.Generator("cpu").manual_seed(0),            
        ).images[0]

image.save("sky_bri_lora_flux_nf4.png")

Trigger words

You should use Sky Bri to trigger the image generation.

Download model

Download them in the Files & versions tab.