search

invoke-ai / InvokeAI

Invoke is a leading creative engine for Stable Diffusion models, empowering professionals, artists, and enthusiasts to generate and create visual media using the latest AI-driven technologies. The solution offers an industry leading WebUI, and serves as the foundation for multiple commercial products.
https://invoke-ai.github.io/InvokeAI/
Apache License 2.0
23.82k stars 2.45k forks source link

[bug]: Model manager can't install Flux LoRa created in OneTrainer #6912

Open Barafu opened 2 months ago

Barafu commented 2 months ago

Is there an existing issue for this problem?

Operating system

Windows

GPU vendor

Nvidia (CUDA)

GPU model

4090

GPU VRAM

No response

Version number

5.0.0.rc1

Browser

Brave

Python dependencies

No response

What happened

When trying to add Flux LoRa that was created using OneTrainer, model manager gives error on import. SDXL LoRa created in OneTrainer works OK.

What you expected to happen

Expected to import the LoRa

How to reproduce the problem

The log says: 

[2024-09-22 10:55:12,952]::[ModelInstallService]::ERROR --> Model install error: I:/training/ready_models/ball_flux.safetensors
InvalidModelConfigException: Unable to determine model type for I:\training\ready_models\ball_flux.safetensors

Additional context

OneTrainer is rather popular, so there will be many published LoRa's created by it.

ComfyUI can use them now, but had this problem too, this is how they fixed it.

Forge can't use them too (no errors whatsoever, but no effect on image). However, Forge can't use many Flux LoRa right now it seems.

Here is an example of such LoRa. ball_flux.zip It is trained on a bunch of blue spheres tagged as "red cube"

Discord username

barafu_albino

Barafu commented 2 months ago

In issue #6916 I created another LoRa that works in in Forge, but can not be imported in Invoke too.

Barafu commented 2 months ago

Still affects 5.0.1