I’m trying to deploy my model to an inference endpoint on AWS. This model works both through the Huggingface UI/API and locally via
diffusers library, both with
accelerate and without, but when trying to deploy it as an inference endpoint I’m getting this weird output: If the logs weren’t truncated to one line I’d be able to tell what are the missing keys, but this way the information is very sparse. Anyone has some idea what could be wrong?
7lprw 2023-08-08T09:30:54.564Z ValueError: Cannot load <class 'diffusers.models.autoencoder_kl.AutoencoderKL'> from /repository/vae because the following keys are missing: 7lprw 2023-08-08T09:30:54.564Z File "/app/huggingface_inference_toolkit/handler.py", line 17, in __init__ 7lprw 2023-08-08T09:30:54.564Z File "/opt/conda/lib/python3.9/site-packages/starlette/routing.py", line 654, in startup 7lprw 2023-08-08T09:30:54.565Z Application startup failed. Exiting.
Here’s the model in question, as you can see it works normally through the web UI. 8clabs/sketch-model-2 · Hugging Face I created this model by importing a safetensors exported from ComfyUI and then exporting it via
Thanks in advance!