Graph sigmas

A schedule you can look at

Schedulers are invisible until the render comes out. Graph sigmas from sigmas_tools_and_the_golden_scheduler fixes that: it plots any SIGMAS tensor as a proper line graph and hands you an IMAGE you can preview or save. You wire in a schedule, a tiny matplotlib chart comes out the other side, and suddenly "Karras vs. exponential vs. this weird merge I just built" is something you can see before spending a minute of GPU time on each.

It's the pack's debugging-and-teaching node, and it pulls its weight in that role. Understanding that a scheduler is just a curve - that Karras pushes effort into the middle while a linear schedule removes equal amounts each step - becomes obvious the first time you graph two of them side by side. If you've been running the sampler/scheduler dropdown without ever seeing what the noise levels actually do, this node is a five-minute education.

What you set

Two inputs in the schema that matters:

• sigmas - any SIGMAS tensor, from a stock scheduler or anything else in this pack
• print_as_list - a boolean, default off. Flip it on and the node prints the raw sigma values to your console, plus each step's value normalized to 0–100%. That's the version of the node for people who want numbers, not a picture

One IMAGE output, which you feed into a standard PreviewImage or SaveImage node. The chart is a line plot with markers, values on the Y axis against step index on the X axis.

Where it shines

Two workflows, mainly. First, the merge debugging loop: build a merged or custom schedule, graph it, notice the shape is wrong, adjust. Second, the "why is this render oversaturated" diagnosis - a schedule that's too aggressive is usually visible as a curve that's too steep too early, before you waste a render confirming it. It also got a nod in community threads about the pack precisely because it makes sigma manipulation legible rather than a mystery box.

Install & gotchas

ComfyUI Manager → search sigmas_tools_and_the_golden_scheduler, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Extraltodeus/sigmas_tools_and_the_golden_scheduler

Restart, no model downloads. This node is the reason the pack needs matplotlib - the graph is drawn with it - and the pack's requirements.txt forgets to list either matplotlib or scipy (it only lists asteval). If the pack errors on import, pip install matplotlib scipy asteval is the fix. Note the output is an IMAGE, so don't try to wire it back into a sampler - the schedule stays on the input side, the picture is just for you.