Save 4 hours! We auto-setup your workflow! Free!

Drop your workflow.json — we handle every dependency, custom node, and model. Just open the link and run.

Auto-Setup Workflow Json (Free) Now!
ComfyUI > Nodes > ComfyUI-SolAttn_triton

ComfyUI Extension: ComfyUI-SolAttn_triton

Repo Name

ComfyUI-SolAttn_triton

Author
kijai (Account age: 0 days)
Nodes
View all nodes(1)
Latest Updated
2026-11-16
Github Stars
0.5K

How to Install ComfyUI-SolAttn_triton

Install this extension via the ComfyUI Manager by searching for ComfyUI-SolAttn_triton
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter ComfyUI-SolAttn_triton in the search bar
After installation, click the Restart button to restart ComfyUI. Then, manually refresh your browser to clear the cache and access the updated list of nodes.

Visit ComfyUI Online for ready-to-use ComfyUI environment

  • Free trial available
  • 16GB VRAM to 80GB VRAM GPU machines
  • 400+ preloaded models/nodes
  • Freedom to upload custom models/nodes
  • 200+ ready-to-run workflows
  • 100% private workspace with up to 200GB storage
  • Dedicated Support

Run ComfyUI Online

ComfyUI-SolAttn_triton Description

ComfyUI-SolAttn_triton uses Triton-based sparse attention to reduce computation in image and video generation, aiming to speed compatible workflows while letting users tune its timing and strength.

ComfyUI-SolAttn_triton Introduction

ComfyUI-SolAttn_triton is deprecated. ComfyUI and comfy-kitchen now include built-in sparse attention, including Sol-Attn. For a new workflow, use ComfyUI’s built-in Model Sparse Attention node with its method set to sol-attn. This community extension remains useful as a reference for older workflows, but its author describes it as redundant. (github.com)

Sol-Attn is a way to reduce some of the work involved in generating images and videos. The extension brought an experimental implementation of that method into ComfyUI using Triton, a tool for running specialized calculations on compatible GPUs. It does not provide a new image or video model, and it does not require retraining your existing model. Its purpose is to help suitable workflows run faster while aiming to preserve visual quality. Results depend on the workflow and settings. (github.com)

The author reported testing the extension with MiniMax H3 on RTX 4090 and RTX 5090 GPUs. That is a limited test record, not a promise that every model or computer will benefit. (github.com)

How ComfyUI-SolAttn_triton Works

During generation, a model repeatedly decides which parts of an image or video should influence other parts. This process is called attention. For a long video, checking every possible relationship can take substantial time.

Think of it like reviewing a large contact sheet. Instead of studying every thumbnail in equal detail, you first glance across the sheet, examine the most relevant areas closely, and use your quick impressions for the rest. Sol-Attn similarly selects important groups for more exact attention calculations and approximates the contribution of other groups. This is called sparse attention. It can reduce computation, though using it more aggressively can increase the chance of visible differences in the result. (arxiv.org)

The extension also lets you choose when sparse attention is used during generation. You can leave some stages using the model’s usual, full attention and apply Sol-Attn only within a chosen portion of the process. (github.com)

ComfyUI-SolAttn_triton Features

  • start_percent and end_percent — choose the active portion of generation. These settings define the window in which the extension applies sparse attention. For example, starting later leaves more of the early generation stages using full attention; ending earlier returns to full attention for the final stages. Adjust the window if you notice changes you do not want in a particular part of the result. (github.com)
  • tau — adjust the speed-versus-quality trade-off. A higher value makes attention sparser, potentially improving speed but increasing the risk of visual differences. A lower value keeps more attention calculations exact. Change it gradually and compare outputs from the same workflow rather than assuming one value is best for every project. (github.com)
  • First-use compilation. The extension compiles its Triton calculations the first time they are used. Your first run may therefore be slower than later runs; it is not a useful run for judging steady-state speed. (github.com)

For visual comparisons, see the author’s test output, the SageAttention example, and the Sol-Attn example. The author also shared this example image. These examples illustrate possible results; they are not a guarantee of speed or quality in your workflow. (github.com)

Troubleshooting ComfyUI-SolAttn_triton

The first generation seems unusually slow.
Let the first run finish, then compare later runs of the same workflow. Triton compilation adds work on first use. (github.com)

The result looks different from your usual output.
Reduce tau to make attention less sparse. You can also start sparse attention later or end it earlier with start_percent and end_percent. Compare the result again using the same model and workflow so the setting you changed is easier to assess. (github.com)

Should you use this extension in a new ComfyUI workflow?
Usually, no. The author marks it deprecated because ComfyUI and comfy-kitchen include Sol-Attn through built-in sparse attention. Check ComfyUI’s MiniMax H3 sparse-attention guide for an artist-oriented example of the built-in approach. (github.com)

Does it include its own generation models?
No models are listed in the extension’s README. It changes how attention is calculated in a compatible generation workflow; it is not a replacement for your image or video model. The author’s stated tests used MiniMax H3. (github.com)

Learn More about ComfyUI-SolAttn_triton

ComfyUI-SolAttn_triton Related Nodes

RunComfy
Copyright 2025 RunComfy. All Rights Reserved.

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.

ComfyUI-SolAttn_triton detailed guide | ComfyUI