Hi, I am trying to use a text2vec-transformer as my vectorizer. I have tried pulling a standard supplied image
- semitechnologies/transformers-inference:distilbert-base-uncased
I have also tried the custom image approach, downloading the models from HuggingFace, and tried using 2 different models. They all resulted in seg faults.
In all cases the endpoint seg faults when I try to POST to the route /vectors. The liveliness + readiness + /meta routes are all good. I can even see the FastApi /docs page.
I am running on Mac OS Silicon chip and verified that the image is for ARM64 architecture.
Ideas? I even tried tags other than latest.