K2 LoRA Info

K2 LoRA Info is the diagnostic node of the K2 Lab suite in comfyui-FVMtools, and it exists because of a very specific, very common failure: you attach a LoRA, it does nothing, and you have no idea why. This node answers that in about a second by reading the file's safetensors header - adapter count, rank, adapter type (lora or lokr), key namespace, and training metadata - without loading a single tensor onto the GPU.

What it checks

The single most useful thing it reports is whether the file even looks like a Krea 2 LoRA, surfaced as the looks_like_krea2 boolean output. The rule is the key namespace: Krea 2 LoRAs live under diffusion_model, blocks, or txtfusion. Anything else - and this is the case that produces "LoRA did nothing" - was trained for a different architecture. It's that simple, and it's cheap to check before you spend a generation.

The report output is the full text dump: adapters, rank, type, namespace, and whatever training metadata the file carries.

When to reach for it

Two times. First, when a regional LoRA seems inert: check the file with this before you start blaming K2 Regional LoRA's routing or the attention router. Second, when you're organizing a Krea 2 LoRA folder and want to know which files are actually usable with this model. Because nothing gets loaded, it's instant and it can't OOM.

It's an output node, so it just runs and shows you the report - no wiring onward required. But there's a neat trick: you can feed looks_like_krea2 into a switch or an if-condition to skip a generation automatically when a file is incompatible. For a batch of LoRAs you're auditioning, that turns a silent failure into a skipped run.

Install

Same as the whole pack:

cd ComfyUI/custom_nodes
git clone https://github.com/ping1979ping/comfyui-FVMtools

Restart ComfyUI, find it under FVM Tools/K2. No extra dependencies. The lora_name dropdown reads from your normal models/loras folder, so it sees exactly the files you'd attach in K2 Regional LoRA.

The companion diagnostic is K2 Plan Report, which tells you whether an installed and correct LoRA actually matched Krea targets in your specific graph. Between the two, "the LoRA did nothing" stops being a mystery and becomes a two-step check: is it a Krea LoRA (LoRA Info), and did it route (Plan Report)?