melMass
comfy_mtb kills WD14 Tagger [bug]
Describe the bug
if i install the MTB nodes my WD14 tagger (https://github.com/pythongosssss/ComfyUI-WD14-Tagger) goes to "IMPORT FAILED" at comfy startup. at the moment i use comfy v0.0.1 but this happened also a version earlier. repairing reinstalling dont help. tagger stayes broken at loadup.
comfy complaining with a few errors at startup. looks like the "onnxruntime_gpu-1.18.1.dist-info" is the problem.
Traceback (most recent call last): File "F:\ComfyUI_v0.0.1_MK3_Testing\ComfyUI\nodes.py", line 1931, in load_custom_node module_spec.loader.exec_module(module) File "
", line 940, in exec_module File " ", line 241, in call_with_frames_removed File "F:\ComfyUI_v0.0.1_MK3_Testing\ComfyUI\custom_nodes\ComfyUI-WD14-Tagger_init.py", line 4, in from .wd14tagger import NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS File "F:\ComfyUI_v0.0.1_MK3_Testing\ComfyUI\custom_nodes\ComfyUI-WD14-Tagger\wd14tagger.py", line 10, in import onnxruntime as ort File "F:\ComfyUI_v0.0.1_MK3_Testing\python_embeded\Lib\site-packages\onnxruntime_init_.py", line 57, in raise import_capi_exception File "F:\ComfyUI_v0.0.1_MK3_Testing\python_embeded\Lib\site-packages\onnxruntime_init_.py", line 23, in from onnxruntime.capi._pybind_state import ExecutionMode # noqa: F401 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "F:\ComfyUI_v0.0.1_MK3_Testing\python_embeded\Lib\site-packages\onnxruntime\capi_pybind_state.py", line 32, in > from .onnxruntime_pybind11_state import * # noqa
internal status (http://127.0.0.1:8188/mtb/status) of MTB tells me this as Failed to load MTB_DeepBump - MTB_FaceSwap - MTB_FilmInterpolation - MTB_ImageRemoveBackgroundRembg - MTB_LoadFaceAnalysisModel - MTB_LoadFaceSwapModel - MTB_LoadFilmModel -
the only way to get both working is to delete the onnxruntime and WD14. after that reinstalling WD14 brings the propper onnxruntime.
Reproduction
No response
Expected behavior
No response
Operating System
Windows (Default)
Comfy Mode
Comfy Portable (embed) (Default)
Console output
No response
Additional context
No response
Thanks for the report, my guess is that you are now using CUDA 12 and it needs a special index to be specified in that case... I'll test a few things to see the best approach here. In the meantime you can follow these instructions: https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-cuda-12x
Using the python/pip from comfy embed (at F:\ComfyUI_v0.0.1_MK3_Testing\python_embeded\bin I think)
thx for having a look. i would guess its the tagger which dont wanna work with the newer/other onnx stuff. maybe it gets an update sometimes.