Open Saganaki22 opened 1 month ago
Can you show the layout of insightface folder? (At a few level if it makes sense) from what I could find online it's either an issue with your onnx version or the layout of the buffalo directory
I have reinstalled both onnxruntime and onnxruntime-gpu to the correct versions outlined in the reactor node troubleshooting
Folder Structure
insightface ├── models │ └── buffalo_l │ ├── 1k3d68.onnx │ ├── 2d106det.onnx │ ├── det_10g.onnx │ ├── genderage.onnx │ ├── inswapper_128.onnx │ └── w600k_r50.onnx
Those are exclusive package(regular and gpu) but it still shouldn't give this kind of error
Those are exclusive package(regular and gpu) but it still shouldn't give this kind of error
what i find odd is that i use insightface in other workflows and i have not had issues like this, i also see the same error with CPU / GPU inference with this workflow
That's indeed weird, I just checked and the layout seems fine. Searching your error gives a lot of hits on the Reactor issue tracker.
And I've seen various errors that can happen with the non gpu version of onnnx but it's not that.
If you are able to run pip freeze
in your comfy python environment it might help spot dependency issues.
Can you confirm that the other workflow were you used insightface still works now too tbs?
That's indeed weird, I just checked and the layout seems fine. Searching your error gives a lot of hits on the Reactor issue tracker.
And I've seen various errors that can happen with the non gpu version of onnnx but it's not that.
If you are able to run
pip freeze
in your comfy python environment it might help spot dependency issues.Can you confirm that the other workflow were you used insightface still works now too tbs?
yes other workflows that use .onnx models and insightface work as of an hr ago, I can use the .pth landmark if need be to bypass the error but i will also do pip freeze to see what may be causing potential conflicts
I have correct insightface, onnx, onnxruntime-gpu version for my python + portable comfyUI version I have buffalo_l and insightface onnx models
I did all the Reactor node troubleshooting
Win10 Portable ComfyUI NVIDIA 3090
still same issue
Anyone got any clues