fermatresearch/dpo-sdxl-controlnet-lora

DPO-SDXL Canny controlnet with LoRA support.

Capabilities

Reference ImagesNegative PromptSeed

Cost

Community model (estimated from hardware time)

Input Parameters

Name	Type	Description	Default	Constraints
`apply_watermark`	boolean	Applies a watermark to enable determining if an image is generated in downstream applications. If you have other provisions for generating or deploying images safely, you can use this to disable watermarking.	`true`	—
`condition_scale`	number	The bigger this number is, the more ControlNet interferes	`0.5`	min: 0, max: 1
`guidance_scale`	number	Scale for classifier-free guidance	`7.5`	min: 1, max: 50
`image`	string(uri)	Input image for img2img or inpaint mode	`—`	—
`lora_scale`	number	LoRA additive scale. Only applicable on trained models.	`0.6`	min: 0, max: 1
`lora_weights`	string	Replicate LoRA weights to use. Leave blank to use the default weights.	`—`	—
`negative_prompt`	string	Input Negative Prompt	`""`	—
`num_inference_steps`	integer	Number of denoising steps	`50`	min: 1, max: 500
`num_outputs`	integer	Number of images to output	`1`	min: 1, max: 4
`prompt`	string	Input prompt	`"An astronaut riding a rainbow unicorn"`	—
`refine`	string	Whether to use refinement steps or not	`"base_image_refiner"`	no_refinerbase_image_refiner
`refine_steps`	integer	For base_image_refiner, the number of steps to refine	`10`	—
`scheduler`	string	scheduler	`"K_EULER"`	DDIMDPMSolverMultistepHeunDiscreteKarrasDPMK_EULER_ANCESTRALK_EULERPNDM
`seed`	integer	Random seed. Leave blank to randomize the seed	`—`	—

apply_watermarkboolean

Applies a watermark to enable determining if an image is generated in downstream applications. If you have other provisions for generating or deploying images safely, you can use this to disable watermarking.

Default: true

condition_scalenumber

The bigger this number is, the more ControlNet interferes

Default: 0.5min: 0, max: 1

guidance_scalenumber

Scale for classifier-free guidance

Default: 7.5min: 1, max: 50

imagestring

Input image for img2img or inpaint mode

lora_scalenumber

LoRA additive scale. Only applicable on trained models.

Default: 0.6min: 0, max: 1

lora_weightsstring

Replicate LoRA weights to use. Leave blank to use the default weights.

negative_promptstring

Input Negative Prompt

Default: ""

num_inference_stepsinteger

Number of denoising steps

Default: 50min: 1, max: 500

num_outputsinteger

Number of images to output

Default: 1min: 1, max: 4

promptstring

Input prompt

Default: "An astronaut riding a rainbow unicorn"

refinestring

Whether to use refinement steps or not

Default: "base_image_refiner"

no_refinerbase_image_refiner

refine_stepsinteger

For base_image_refiner, the number of steps to refine

Default: 10

schedulerstring

scheduler

Default: "K_EULER"

DDIMDPMSolverMultistepHeunDiscreteKarrasDPMK_EULER_ANCESTRALK_EULERPNDM

seedinteger

Random seed. Leave blank to randomize the seed

Version: 5779933f36e3Updated: 7/25/2026795 runs