This message was deleted.

Hey! That usually means that your service code has some errors in it. Can you run it with python like

python service.py

?

Could you please share your service code here?

import bentoml from bentoml.io import JSON model_ref = bentoml.xgboost.get("credit_risk_model:icfexlspfwvdaacc") model_runner = model_ref.to_runner() svc = bentoml.Service("credit_risk_classifier", runners=[model_runner]) @svc.api(input=JSON(),output=JSON()) def classify(application_data): prediction = model_runner.predict.run(application_data) return {"status": "Approved"}

Oh! I just realized the actual issue, you need to put

service.py:svc

and not

service:svc

, I think!

I used bentoml serve service.py:svc and still has this error: [bentoml-cli]

serve

failed: Failed to load bento or import service 'service.py:svc'. If you are attempting to import bento in local store:

Attribute "svc" not found in module "service".

, or if you are importing by python module path:

Bento 'service.py:svc' is not found in BentoML store <osfs 'C:\Users\yadwi\bentoml\bentos'>

Oh, you're passing it on the command line! It should just be

bentoml serve .

, then. For the second error, it seems like your model hasn't saved the custom object properly, can you

print(model_ref.custom_objects)

?

I'm using anaconda prompt and print('model_ref.custom_objects) gives Unable to initialize device PRN

Ah, that should be inside the service code, right above the line where you try to use

dictVectorizer

, not in the prompt. You can just run the service again after to see the output.

tried it, still the same error.....

Yep, it should still error, but it should print the value of custom objects now.

Okay, so it worked using this solution: https://bentoml.slack.com/archives/CKRANBHPH/p1666284917697239?thread_ts=1666247662.933749&cid=CKRANBHPH

Ah, great that you worked that out!

Thanks for all the support and prompt responses 🙂