Awesome

ComfyUI_TGate

ComfyUI reference implementation for T-GATE.

T-GATE could brings 10%-50% speed up for different diffusion models, only slightly reduces the quality of the generated images and maintains the original composition.

Some monkey patch is used for current implementation. If any error occurs, make sure you have the latest version.

If my work helps you, consider giving it a star.

Some of my other projects that may help you.

• ComfyUI-TCD
• ComfyUI_TGate
• ComfyUI-ELLA

:star2: Changelog

• [2024.5.23]: updated to latest ComfyUI version
• [2024.5.15]:
  • Fix sdxl x, y batch error. If you encounter some errors, you can try TGate Apply(Deprecated) node.
• [2024.5.06]:
  • Add use_cpu_cache, reduce some GPU OOM promblem.
  • Refactor apply node, Deprecated old monkey patching method node. The legacy one will be removed after few version.
  • Add a simple node and advanced node.
• [2024.4.30] :wrench: Fixed an cond-only sampling bug that caused animatediff error. Thanks pamparamm.
• [2024.4.29] :wrench: TL,DR: Improved performance and T-GATE only works where it needs to work.
  • Fixed a bug that caused TGateApply to affect other places where the model is used, even if TGateApply is turned off.
  • Fixed cross attntion results not being cached correctly causing performance to be slightly lower than the git patch version.
• [2024.4.26] :tada: Native version release(NO NEED git patch anymore!).
• [2024.4.18] Initial repo.

:books: Example workflows

The examples directory has workflow example. There are images generated with and without T-GATE in the assets folder.

Origin result	T-GATE result

T-GATE result image comes from the workflow included in the example image.

Compare to AutomaticCFG

AutomaticCFG is another ComfyUI plugin: Your CFG won't be your CFG anymore. It is turned into a way to guide the CFG/final intensity/brightness/saturation, and it adds a 30% speed increase.

env: T4-8G

	Origin	T-GATE 0.5	AutomaticCFG	T-GATE 0.35	AutomaticCFG fatest
result
speed	4.59it/s	5.68it/s	5.62it/s	6.13it/s	6.13it/s

T-GATE performs best when maintaining the original composition. However, if you don't need to maintain composition, AutomaticCFG fatest also brings about the same performance improvement.

:green_book: INSTALL

git clone https://github.com/JettHu/ComfyUI_TGate
# that's all!

:orange_book: Major Features

• Training-Free.
• Friendly support CNN-based U-Net, Transformer, and Consistency Model
• 10%-50% speed up for different diffusion models.

:book: Nodes reference

TGate Apply

Inputs

• model, model loaded by Load Checkpoint and other MODEL loaders.

Configuration parameters

• start_at, this is the percentage of steps. Defines at what percentage point of the generation to start use the T-GATE cache.
• use_cpu_cache: If multiple batches (animatediff) cause GPU OOM, you can set it to true, and T-GATE performance will decrease.

TGate Apply Advanced

Inputs

• model, model loaded by Load Checkpoint and other MODEL loaders.

Configuration parameters

• start_at, this is the percentage of steps. Defines at what percentage point of the generation to start use the T-GATE cache.
• only_cross_attention, [RECOMMEND] default is True, the effect is to cache only the output of cross-attention, ref to issues
• use_cpu_cache: If multiple batches (animatediff) cause GPU OOM, you can set it to true, and T-GATE performance will decrease.

Optional configuration

• self_attn_start_at, only takes effect when only_cross_attention is false, percentage of steps too. Defines at what percentage point of the generation to start use the T-GATE cache on latent self attnention.

TGate Apply(Deprecated)

This node is already deprecated, and will be removed after few version.

Inputs

• model, model loaded by Load Checkpoint and other MODEL loaders.

Configuration parameters

• start_at, this is the percentage of steps. Defines at what percentage point of the generation to start use the T-GATE cache.
• only_cross_attention, [RECOMMEND] default is True, the effect is to cache only the output of cross-attention, ref to issues
• use_cpu_cache: If multiple batches (animatediff) cause GPU OOM, you can set it to true, and T-GATE performance will decrease.

Optional configuration

• self_attn_start_at, only takes effect when only_cross_attention is false, percentage of steps too. Defines at what percentage point of the generation to start use the T-GATE cache on latent self attnention.

:rocket: Performance (from T-GATE)

The latency is tested on a 1080ti commercial card.

The MACs and Params are calculated by calflops.

The FID is calculated by PytorchFID.

:memo: TODO

• Result image quality is inconsistent with origin. Now cache attn2 (cross_attention) only.
• Implement a native version and no longer rely on git patch
• Fully compatible with animatediff. Currently, both plugins hook comfy.samplers.sampling_function, T-Gate does not perform correctly. refer to
• compatible with TiledDiffusion. issue #11

:mag: Common promblem

• For apple silicon users using the mps backend, torch and macos versions may cause some problems. refer to issue comment.

• Fixed in 2024.4.29. Unable to properly remove T-Gate effects. The situation in the picture below is bypass the node after apply.

2024.4.26-29	Updated on 2024.4.29