module transformers has no attribute CLIPImageProcessor

#6
by siemens - opened

when running at pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16) have error "module transformers has no attribute CLIPImageProcessor"

i'm in new transformers4.24.0 diffusers 0.7.2 pytorch 1.10.2

image.png

ran into this same issue

Not 100% sure where this comes from. I will convert the model again and retry it. If that doesn't work I'll retrain the model and update it here.
Thanks for the reports and your patience!

If it helps, I've been able to convert every model of yours to ONNX except this one. When I try, I get a similar attribute error:
File "...\diffusers\src\diffusers\pipeline_utils.py", line 192, in from_pretrained is_pipeline_module = hasattr(pipelines, library_name) TypeError: hasattr(): attribute name must be string

I'll try to push an update to the diffusers to see if it helps : https://huggingface.co/nitrosocke/Nitro-Diffusion/discussions/8
waiting on Nitro to check and merge

I'm further along with the latest commit, but I think it's an ONNX problem now:
onnxruntime.capi.onnxruntime_pybind11_state.NotImplemented: [ONNXRuntimeError] : 9 : NOT_IMPLEMENTED : Could not find an implementation for ArgMax(13) node with name 'ArgMax_1707'

Nevermind, that seems to be a CPU only issue. It's working correctly with DML.

Perfect, glad it's working now, thanks for the fix @Guizmus

nitrosocke changed discussion status to closed

Sign up or log in to comment