Updated feature-extraction API URL
Hello!
If you were using the undocumented feature-extraction API until now, then you may have noticed that it stopped working. Instead, the API URL has moved to the following URL for this model:
For example:
curl -L \
-X POST \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $HF_API_KEY" \
/static-proxy?url=https%3A%2F%2Frouter.huggingface.co%2Fhf-inference%2Fmodels%2Fsentence-transformers%2Fall-MiniLM-L6-v2%2Fpipeline%2Ffeature-extraction \
-d '{"inputs": ["hello"] }'
And in addition, this URL can be used for the sentence-similarity task for this model:
Apologies for the inconvenience.
- Tom Aarsen
I've been using this API without issues until today, but as of today every request I send returns a server error as outlined here /static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fapi-error-for-model-sentence-transformers-all-minilm-l6-v2%2F168083%3C%2Fa%3E%3C%2Fp%3E
I've also been using the API URL for a separate module in my app that's used to generate embeddings to upsert content where I've been using the URL /static-proxy?url=https%3A%2F%2Fapi-inference.huggingface.co%2Fmodels%2Fsentence-transformers%2Fall-MiniLM-L6-v2%2Fpipeline%2Ffeature-extraction%3C%2Fa%3E and as of today it stopped working. I also tried to change this URL to /static-proxy?url=https%3A%2F%2Frouter.huggingface.co%2Fhf-inference%2Fmodels%2Fsentence-transformers%2Fall-MiniLM-L6-v2%2Fpipeline%2Ffeature-extraction%3C%2Fa%3E but it still returns a 504 error. Is something going on?
For me for now on Web:
/static-proxy?url=https%3A%2F%2Fapi-inference.huggingface.co%2Fmodels%2Fsentence-transformers%2Fall-MiniLM-L6-v2%2Fpipeline%2Ffeature-extraction%3C%2Fa%3E # 404
/static-proxy?url=https%3A%2F%2Frouter.huggingface.co%2Fhf-inference%2Fmodels%2Fsentence-transformers%2Fall-MiniLM-L6-v2%2Fpipeline%2Ffeature-extraction%3C%2Fa%3E # Invalid username or password.
Hello!
I'm having a look now. My understanding is that the issues have been resolved as of about 4 hours ago. I'm seeing successful requests coming in again, and I'm also able to use the widget in the model card again.
Apologies for the inconvenience.
- Tom Aarsen
Thank you, tomaarsen!
I'm still facing an issue.
Noted, thanks for reporting. Will push this up the stream and get it sorted.
- Tom Aarsen
I increased the number of instances hosting the model, but latency is still reported as being extremely high. It's been reported in the infrastructure team.
- Tom Aarsen
Thanks, tomaarsen
I increased the number of instances hosting the model, but latency is still reported as being extremely high. It's been reported in the infrastructure team.
- Tom Aarsen
I just checked my logs and it looks like everything was working fine briefly, but then it went back to failing, and now all embeddings are failing at the moment as far as I can tell.
I finally this new url but not sure if I'm using it correctly. Cannot get rid of this error.
Error:
WARNING:retry.api:The model is currently loading, please re-run the query., retrying in 10 seconds...
WARNING:retry.api:The model is currently loading, please re-run the query., retrying in 10 seconds...
RuntimeError Traceback (most recent call last)
/tmp/ipython-input-2136977844.py in <cell line: 0>()
13 "Should I sign up for Medicare Part B if I have Veterans’ Benefits?"]
14
---> 15 output = query(texts)
in query(texts)