nvidia--tensorrt/demo/Diffusion/docs/support_matrix.md

Supported Diffusion Models

This demo supports Diffusion models that are popular in the Generative AI community. The table below lists the various configurations we support for each pipeline.

Pipeline Support Matrix

Pipeline	Version	Task	Supported Precisions	Additional features	Hub	Restrictions
Stable Diffusion	1.4	Text-to-image Image-to-image	FP16, FP8, INT8 *	N/A	CompVis/stable-diffusion-v1-4
Stable Diffusion	dreamshaper-7	Text-to-image Image-to-image	FP16	N/A	Lykon/dreamshaper-7
Stable Diffusion	XL 1.0-base	Text-to-image Image-to-image	FP16, FP8, INT8 *	LoRA (FP16, BF16, FP8)	stabilityai/stable-diffusion-xl-base-1.0
Stable Diffusion	XL 1.0-refiner	Text-to-image Image-to-image	FP16	N/A	stabilityai/stable-diffusion-xl-refiner-1.0
Stable Diffusion	XL-Turbo	Text-to-image Image-to-image	FP16	N/A	stabilityai/sdxl-turbo
Stable Diffusion	3	Text-to-image	FP16	N/A	stabilityai/stable-diffusion-3-medium
Stable Diffusion	3.5-medium	Text-to-image	FP16, BF16	N/A	stabilityai/stable-diffusion-3-medium
Stable Diffusion	3.5-large	Text-to-image	FP16, BF16, FP8	N/A	stabilityai/stable-diffusion-3-large
ControlNet	1.4	Image-to-image	FP16	N/A	lllyasviel/sd-controlnet-canny lllyasviel/sd-controlnet-depth lllyasviel/sd-controlnet-hed lllyasviel/sd-controlnet-mlsd lllyasviel/sd-controlnet-normal lllyasviel/sd-controlnet_openpose lllyasviel/sd-controlnet_scribble lllyasviel/sd-controlnet_seg
ControlNet	XL 1.0-base	Image-to-image	FP16, FP8	N/A	stabilityai/controlnet-canny-sdxl-1.0
ControlNet	3.5-large	Image-to-image	FP16, BF16, FP8 (canny and depth only)	N/A	stabilityai/stable-diffusion-3.5-large-controlnet-canny stabilityai/stable-diffusion-3.5-large-controlnet-depth stabilityai/stable-diffusion-3.5-large-controlnet-blur
Stable Video Diffusion	XT-1.1	Text-to-video	FP16, FP8	N/A	stabilityai/stable-video-diffusion-img2vid-xt-1-1
Stable Cascade	N/A	Text-to-image	BF16	N/A	stabilityai/stable-cascade-prior stabilityai/stable-cascade
Flux	1-Dev	Text-to-image Image-to-image	FP16, BF16, FP8, FP4 *	LoRA (FP16, BF16, FP8)	black-forest-labs/FLUX.1-dev
Flux	1-Schnell	Text-to-image Image-to-image	FP16, BF16, FP8, FP4 *	LoRA (FP16, BF16, FP8)	black-forest-labs/FLUX.1-schnell
Flux	1-Canny-Dev	Image-to-image	FP16, BF16, FP8, FP4	N/A	black-forest-labs/FLUX.1-Canny-dev
Flux	1-Depth-Dev	Image-to-image	FP16, BF16, FP8, FP4	N/A	black-forest-labs/FLUX.1-Depth-dev
Flux	1-Kontext-Dev	Image-to-image	BF16, FP8, FP4	N/A	black-forest-labs/FLUX.1-Kontext-dev
Cosmos	cosmos-predict2-2b-text2image, cosmos-predict2-14b-text2image	Text-to-image	BF16	N/A	nvidia/Cosmos-Predict2-2B-Text2Image nvidia/Cosmos-Predict2-14B-Text2Image
Cosmos	cosmos-predict2-2b-video2world, cosmos-predict2-14b-video2world	Video-to-World	BF16	N/A	nvidia/Cosmos-Predict2-2B-Video2World nvidia/Cosmos-Predict2-14B-Video2World
Wan	2.2-T2V-A14B	Text-to-video	BF16	N/A	wan-ai/wan2.2-t2v-a14b

*Note: Only the text2image pipelines support FP4/FP8/INT8 quantization. The image2image pipelines don't support quantization.