MinusZone - TrainToolsDebug

MZ_TrainToolsDebug is the debugging escape hatch this pack ships for itself - and honestly, for you. This pack's whole architecture is built on passing around opaque custom objects: workspace_config, train_config, advanced_config are all dicts wearing custom types, and ComfyUI shows you nothing about their contents. When a train run behaves mysteriously, this node is how you look inside.

It takes any object (* type - the AlwaysEqualProxy trick means it accepts literally anything and doesn't care what type the other end claims) and pretty-prints it as formatted text, Python's pprint under the hood. Output is a STRING called debug, which you can show with a text display node or just read from the node's output.

The inputs that matter

• object - whatever you want to inspect. Wire the workspace_config from an InitWorkspace node, the train_config from MZ_KohyaSSUseConfig, or the advanced_config from either AdvConfig node. Because it accepts anything, you can also use it on plain strings and numbers to see what actually flowed through.
• index - the power feature. A dot-separated path into the object: "train_config.learning_rate" prints just that field, "0.lora_name" digs into a list/dict at key 0. Great for confirming one value without drowning in the whole dump.
• indent, depth, width, sort_keys, underscore_numbers - pprint formatting knobs. The defaults (2, 5, 80, sorted, underscored) are fine; depth is the one to raise if a nested config prints as ....

How to use it

The classic debugging loop with this pack: something's off in training → wire MZ_TrainToolsDebug onto the train_config (or the workspace_config) → run the graph → read the dump → find the field that's empty or wrong (the classic: dataset_config pointing at a stale path, or network_weights set when you wanted a fresh run). Then fix it upstream and remove the debug node. It's a diagnostic tool, not a permanent graph citizen, though plenty of people leave it hanging off the config nodes permanently - it's cheap and it's the only window into these custom types.

Install

Same as everything else in the pack:

cd ComfyUI/custom_nodes
git clone https://github.com/MinusZoneAI/ComfyUI-TrainTools-MZ
# restart ComfyUI

Or ComfyUI Manager → search ComfyUI-TrainTools-MZ.

A note on expectations: this node shows you the config the pack built, but it won't tell you whether that config is right for kohya - for that you compare against sd-scripts docs or the template JSONs in the pack's configs/kohya_ss_lora/ folder, which are the ground truth for what fields the trainer actually reads. And because the pack runs training in-process, a debug node won't catch kohya's own argument errors - those still land in the ComfyUI console. Use this node for "what is the pack about to do," and the console for "why did it blow up."