__class__ not set defining 'ModelSettings' error during mlserver start

[sabaimran]

Hi all! I wanted to get a quick start example going to see if mlserver was a good candidate for my use case. I ran into this issue when trying to start the server:

cmd: mlserver start .
This is the error log:

Traceback (most recent call last):
  File "/<path>/.venv/bin/mlserver", line 5, in <module>
    from mlserver.cli import main
  File "/<path>/.venv/lib/python3.10/site-packages/mlserver/__init__.py", line 2, in <module>
    from .server import MLServer
  File "/<path>/.venv/lib/python3.10/site-packages/mlserver/server.py", line 7, in <module>
    from .model import MLModel
  File "/<path>/.venv/lib/python3.10/site-packages/mlserver/model.py", line 3, in <module>
    from .codecs import (
  File "/<path>/.venv/lib/python3.10/site-packages/mlserver/codecs/__init__.py", line 1, in <module>
    from .numpy import NumpyCodec, NumpyRequestCodec
  File "/<path>//.venv/lib/python3.10/site-packages/mlserver/codecs/numpy.py", line 7, in <module>
    from .base import InputCodec, register_input_codec, register_request_codec
  File "/<path>/.venv/lib/python3.10/site-packages/mlserver/codecs/base.py", line 6, in <module>
    from ..logging import logger
  File "/<path>/.venv/lib/python3.10/site-packages/mlserver/logging.py", line 10, in <module>
    from .settings import Settings
  File "/<path>/.venv/lib/python3.10/site-packages/mlserver/settings.py", line 192, in <module>
    class ModelSettings(BaseSettings):
RuntimeError: __class__ not set defining 'ModelSettings' as <class 'mlserver.settings.ModelSettings'>. Was __classcell__ propagated to type.__new__?

Files

./settings.json

{
    "host": "0.0.0.0",
    "http_port": 5000,
    "grpc_port": 5001,
    "debug": true
}

./test-model/model-settings.json

{
    "name": "wine-classifier",
    "implementation": "mlserver_mlflow.MLflowRuntime",
    "parameters": {
        "uri": "s3://ml-training-artifacts-pg/1/model"
    }
}

[adriangonz]

Hey @sabaimran ,

That's quite a strange error - I've never experienced that one myself (nor seen that reported). It could be due to a weird mix of dependency versions?

Is this a fresh environment? Could you share the output of pip freeze?

[sabaimran]

Hi @adriangonz! I had other dependencies installed as well. I ended up uninstalling mlserver, but I will try to recreate the environment and see if I can get a reproduction of the error.

[adriangonz]

Hey @sabaimran ,

From what was shared in #1010, this issue may have been caused by an old version of Pydantic (<= 1.7). If you upgrade to a more recent version, it should go away.

I'll close this for now, but please do re-open if the error re-occurs after upgrading.