lucataco / flux.1-turbo-alpha

8-step distilled lora for FLUX.1-dev model released by the Alimama-Creative Team

  • Public
  • 1.9K runs
  • A100 (80GB)
  • GitHub
  • Weights
  • License

Input

*string
Shift + Return to add a new line

Prompt for generated image

string

Aspect ratio for the generated image. The size will always be 1 megapixel, i.e. 1024x1024 if aspect ratio is 1:1. To use arbitrary width and height, set aspect ratio to 'custom'.

Default: "1:1"

integer
(minimum: 256, maximum: 1440)

Width of the generated image. Optional, only used when aspect_ratio=custom. Must be a multiple of 16 (if it's not, it will be rounded to nearest multiple of 16)

integer
(minimum: 256, maximum: 1440)

Height of the generated image. Optional, only used when aspect_ratio=custom. Must be a multiple of 16 (if it's not, it will be rounded to nearest multiple of 16)

integer
(minimum: 1, maximum: 4)

Number of images to output.

Default: 1

integer
(minimum: 1, maximum: 30)

Number of inference steps

Default: 8

number
(minimum: 0, maximum: 10)

Guidance scale for the diffusion process

Default: 3.5

integer

Random seed. Set for reproducible generation

string

Format of the output images

Default: "webp"

integer
(minimum: 0, maximum: 100)

Quality when saving the output images, from 0 to 100. 100 is best quality, 0 is lowest quality. Not relevant for .png outputs

Default: 80

boolean

This model’s safety checker can’t be disabled when running on the website. Learn more about platform safety on Replicate.

Disable safety checker for generated images. This feature is only available through the API. See [https://replicate.com/docs/how-does-replicate-work#safety](https://replicate.com/docs/how-does-replicate-work#safety)

Default: false

Output

output
Generated in

Run time and cost

This model costs approximately $0.13 to run on Replicate, or 7 runs per $1, but this varies depending on your inputs. It is also open source and you can run it on your own computer with Docker.

This model runs on Nvidia A100 (80GB) GPU hardware. Predictions typically complete within 95 seconds. The predict time for this model varies significantly based on the inputs.

Readme

About

This repository provides a 8-step distilled lora for FLUX.1-dev model released by AlimamaCreative Team.

Description

This checkpoint is a 8-step distilled Lora, trained based on FLUX.1-dev model. We use a multi-head discriminator to improve the distill quality. Our model can be used for T2I, inpainting controlnet and other FLUX related models. The recommended guidance_scale=3.5 and lora_scale=1. Our Lower steps version will release later.

  • Text-to-Image.

How to use

diffusers

This model can be used directly with diffusers

comfyui

Training Details

The model is trained on 1M open source and internal sources images, with the aesthetic 6.3+ and resolution greater than 800. We use adversarial training to improve the quality. Our method fix the original FLUX.1-dev transformer as the discriminator backbone, and add multi heads to every transformer layer. We fix the guidance scale as 3.5 during training, and use the time shift as 3.

Mixed precision: bf16

Learning rate: 2e-5

Batch size: 64

Image size: 1024x1024

Licensing and commercial use

If you generate images on Replicate with FLUX.1 models and their fine-tunes, then you can use the images commercially.

If you download the weights off Replicate and generate images on your own computer, you can’t use the images commercially.