I think `flytekitplugins pydantic` is incompatible with `pyd Flyte #ask-the-community

I think `flytekitplugins-pydantic` is incompatible...

Kirill Dubovikov

02/26/2024, 3:15 PM

I think

flytekitplugins-pydantic

is incompatible with

pydantic 2.x

. Whenever I try to run my code I get errors like this

Copy code

File "/Users/kdubovikov/Library/Caches/pypoetry/virtualenvs/universal-model-flyte--i4p14Mq-py3.11/lib/python3.11/site-packages/flytekitplugins/pydantic/__init__.py", line 1, in <module>
    from .basemodel_transformer import BaseModelTransformer
  File "/Users/kdubovikov/Library/Caches/pypoetry/virtualenvs/universal-model-flyte--i4p14Mq-py3.11/lib/python3.11/site-packages/flytekitplugins/pydantic/basemodel_transformer.py", line 12, in <module>
    from . import deserialization, serialization
  File "/Users/kdubovikov/Library/Caches/pypoetry/virtualenvs/universal-model-flyte--i4p14Mq-py3.11/lib/python3.11/site-packages/flytekitplugins/pydantic/deserialization.py", line 14, in <module>
    PYDANTIC_VALIDATOR_METHOD_NAME = pydantic.BaseModel.__get_validators__.__name__
                                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/kdubovikov/Library/Caches/pypoetry/virtualenvs/universal-model-flyte--i4p14Mq-py3.11/lib/python3.11/site-packages/pydantic/_internal/_model_construction.py", line 221, in __getattr__
    raise AttributeError(item)
AttributeError: __get_validators__

According to the docs,

__get_validators__

should be replaced with `__get_pydantic_core_schema__`: https://docs.pydantic.dev/latest/migration/#defining-custom-types

Ketan (kumare3)

02/26/2024, 3:31 PM

I thought someone was working on pydantic 2.0 support. Cc @Thomas Fan or @L godlike

Kirill Dubovikov

02/26/2024, 3:31 PM

It may be good to note that 2.x is not supported in the docs

L godlike

02/26/2024, 3:50 PM

Can take a look tommorow, thank you

L godlike

02/27/2024, 11:55 AM

Copy code

from flytekit import task
# from pydantic import BaseModel
from pydantic.v1 import BaseModel


class TrainConfig(BaseModel):
    lr: float = 1e-3
    batch_size: int = 32
    # files: List[FlyteFile]
    # directories: List[FlyteDirectory]

@task
def train() -> TrainConfig:
    return cfg

L godlike

02/27/2024, 11:55 AM

I just run this example with

Copy code

(dev) future@outlier ~ % pip list | grep -i pydantic
autodoc-pydantic                         1.8.0
pydantic                                 2.6.2
pydantic_core                            2.16.3

L godlike

02/27/2024, 1:09 PM

with this image

Copy code

futureoutlier/flytekit:pydantic-v3

L godlike

02/27/2024, 1:09 PM

fixed!

L godlike

02/27/2024, 2:53 PM

This pr here can help you. https://github.com/flyteorg/flytekit/pull/2217

2 Views

Open in Slack

Previous Next