Trained model (.tdict) produces black image in Text to Image — v2.5.1

I trained a custom model using DiffusionBee's built-in training feature (Dreambooth/LoRA, ~10 images of a person). Training completed and produced a .tdict file. The model appears in My Models and is selectable in the Text to Image model dropdown. However, every generation produces a solid black image.
The default DiffusionBee model generates images normally with the same settings.
Steps to reproduce:

Train a custom model using built-in training
Select trained .tdict model in Text to Image
Generate with any prompt including trigger word
Result: solid black image

Environment:

DiffusionBee version: 2.5.1
Mac: MacBook Air M2
macOS version: Sequoia 15.4.1

What I've tried:

LoRA slot shows "None" with no other options — cannot load model as add-on
Default model works fine
Adjusting steps, CFG scale, sampler — no change