GPU Status Display (Debug)

Let's be blunt about what this node is: GPUStatusDisplayDebug is a subclass of GPU Status Display with the category changed from MultiGPU/Diagnostics to MultiGPU/Debug. Same inputs, same outputs, same code path. It's not a "more detailed" debug build - the pack author just re-registered the same node under the Debug menu so you can keep your debugging graphs together.

Which sounds pointless until you're building the exact workflow this pack's USER_GUIDE recommends: a debug loader → debug sampler → display node chain, all living under MultiGPU/Debug. If your graph is already swimming in debug nodes, having the status display in the same menu means you're not hunting through Diagnostics mid-problem. That's the whole value, and it's a small one.

What it does

Same as the parent: reads every CUDA device via torch.cuda, reports used/total VRAM per GPU as text, and draws a bar-chart image into status_image. Feed it the model output of a MultiGPU loader and it appends the shard layout (input_blocks.0 → cuda:0, etc.). All three inputs (model, status, layout) are optional - it runs fine bare. Outputs are status_text (STRING) and status_image (IMAGE).

For the workflow the guide pushes: connect the sampler's status STRING into this node's status input, and its status_text into a display node, and you get the full per-GPU story in comfyui.log after a run - shard counts, VRAM deltas, peaks.

Installing it

It comes with the Comfy-MultiGPU-Loader pack. Install via ComfyUI Manager (search "Comfy-MultiGPU-Loader") or:

cd ComfyUI/custom_nodes
git clone https://github.com/AngelCookiesLab/Comfy-MultiGPU-Loader
pip install -r requirements.txt

Restart ComfyUI, look under MultiGPU/Debug.

One honest note

Since it's the identical node, if you're not running a debug graph there's zero reason to pick this over the base version. If you are, it's a convenience, not a capability. And remember the project status: this pack is sunset/"as-is" (GPL-3.0, no active support), and the author flags that diagnostics can under-report activity on later runs - so treat its status text as "best effort," not gospel.