Cached model: URL persists in Edit Endpoint page even after clearing and saving
Tried the model caching feature by adding a huggingface URL by editing a pre-existing serverless endpoint. I input the full URL instead of the org/model name format as per docs along with a HF token since it was a gated model.
I started getting ⚡｜serverless"model not found" and initialization hangs errors (which I didn't notice at the time) and then I decided to experiment with downloading the model directly during the docker build process. I couldn't properly deploy this new version because of the model not found errors, which left all workers stuck in
I then tried to remove the model URL from the
I then removed a 24GB GPU configuration for the sake of another change in the config, removed the URL again and saved the changes. This made the
---
Endpoint ID:
---
Edit: 2 of the Extra workers now say
I started getting ⚡｜serverless"model not found" and initialization hangs errors (which I didn't notice at the time) and then I decided to experiment with downloading the model directly during the docker build process. I couldn't properly deploy this new version because of the model not found errors, which left all workers stuck in
initializingI then tried to remove the model URL from the
Edit Endpointxmodel not foundEdit EndpointI then removed a 24GB GPU configuration for the sake of another change in the config, removed the URL again and saved the changes. This made the
model not found---
Endpoint ID:
pyj74rjjbt46dl---
Edit: 2 of the Extra workers now say
model not foundidleinitializingmodel not found
