Closed rbavery closed 1 month ago
rbavery
commented
8 months ago downgrading to provided a more informative traceback that showed I was missing Optional type annotations on some fields.
• Downgrading pydantic-core (2.16.3 -> 2.6.3)
• Downgrading pydantic (2.6.2 -> 2.3.0)When I addressed these issues I could run model_validate_json but only with pydantic 2.3. I get the same error with the Optional typing issues addressed when I tried to use version 2.6.3 again.
sydney-runkle
commented
8 months ago @rbavery,
Could you provide a MRE for this issue? I'm not able to view the commit that you linked with the model. A boiled down version of this issue will help us to address your problem! Thanks.
rbavery
commented
8 months ago For me this week, making an MRE would be too large of a project, I'm not sure how long it would take to reproduce the issue without the library I'm using. Since I don't know what causes the issue in the first place (there's no information in the traceback about where the issue occurred).
I just noticed the formatting was off in my original post, sorry about that.
My package source is here: https://github.com/rbavery/dlm-extension/tree/validate/stac_model
and it is pip installable and testable with the code provided above
the package I'm working with can be installed with
pip install stac-model==0.1.1.alpha3then the error can be reproduced with
from stac_model.examples import eurosat_resnet from stac_model.schema import MLModel model_metadata = eurosat_resnet() metadata_json = model_metadata.model_dump_json(indent=2) model_metadata_validated = MLModel.model_validate_json(metadata_json)
sydney-runkle
commented
8 months ago @rbavery,
If we don't have a purely pydantic MRE, I don't think it makes sense to have this issue open here. I'd recommend opening an issue on the stac_model repo!
Thanks!
rbavery
commented
8 months ago @sydney-runkle I'm the maintainer of stac-model, and the primary dependency is pydantic.
https://github.com/rbavery/dlm-extension/blob/validate/pyproject.toml#L58
rbavery
commented
8 months ago @sydney-runkle it actually looks like I provided a minimal repro back in December, sorry for forgetting to raise this in a new issue: https://github.com/pydantic/pydantic/issues/8189#issuecomment-1848226486
I just tested this with pydantic 2.3.2 and the issue still occurs
# rave at rave-desktop in ~/work/dlm-extension on git:validate ✖︎ [11:21:33]
→ pip install pydantic=="2.6.3"
Collecting pydantic==2.6.3
Obtaining dependency information for pydantic==2.6.3 from https://files.pythonhosted.org/packages/ac/86/c98520827f58c8753783be4bf2286b4f73a18ac71c93ab597ae1aeb26fc8/pydantic-2.6.3-py3-none-any.whl.metadata
Downloading pydantic-2.6.3-py3-none-any.whl.metadata (84 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 84.4/84.4 kB 1.4 MB/s eta 0:00:00
Requirement already satisfied: annotated-types>=0.4.0 in /home/rave/miniforge3/lib/python3.10/site-packages (from pydantic==2.6.3) (0.6.0)
Requirement already satisfied: pydantic-core==2.16.3 in /home/rave/miniforge3/lib/python3.10/site-packages (from pydantic==2.6.3) (2.16.3)
Requirement already satisfied: typing-extensions>=4.6.1 in /home/rave/.local/lib/python3.10/site-packages (from pydantic==2.6.3) (4.8.0)
Downloading pydantic-2.6.3-py3-none-any.whl (395 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 395.2/395.2 kB 3.1 MB/s eta 0:00:00
Installing collected packages: pydantic
Attempting uninstall: pydantic
Found existing installation: pydantic 2.3.0
Uninstalling pydantic-2.3.0:
Successfully uninstalled pydantic-2.3.0
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
fastapi 0.104.1 requires anyio<4.0.0,>=3.7.1, but you have anyio 4.2.0 which is incompatible.
lightly 1.4.23 requires pydantic<2,>=1.10.5, but you have pydantic 2.6.3 which is incompatible.
Successfully installed pydantic-2.6.2
(base)
# rave at rave-desktop in ~/work/dlm-extension on git:validate ✖︎ [11:24:18]
→ ipython
Python 3.10.12 | packaged by conda-forge | (main, Jun 23 2023, 22:40:32) [GCC 12.3.0]
Type 'copyright', 'credits' or 'license' for more information
IPython 8.18.1 -- An enhanced Interactive Python. Type '?' for help.
...: def validate_key(cls, v):
...: if '//' in v:
...: raise ValueError('Key must not contain double slashes')
...: return v.strip('/')
...:
...: class ModelArtifact(BaseModel):
...: path: S3Path | FilePath | str = Field(...)
...: additional_files: Optional[Dict[str, FilePath]] = None
...:
...: class Config:
...: arbitrary_types_allowed = True
...:
...: class ClassMap(BaseModel):
...: class_to_label_id: Dict[str, int]
...:
...: @property
...: def label_id_to_class(self) -> Dict[int, str]:
...: return {v: k for k, v in self.class_to_label_id.items()}
...:
...: class ModelMetadata(BaseModel):
...: signatures: ModelSignature
...: artifact: ModelArtifact
...: id: str = Field(default_factory=lambda: uuid4().hex)
...: class_map: ClassMap
...: runtime_config: Optional[RuntimeConfig] = None
...: name: str
...: ml_model_type: Optional[str] = None
...: ml_model_processor_type: Optional[Literal["cpu", "gpu", "tpu", "mps"]] = None
...: ml_model_learning_approach: Optional[str] = None
...: ml_model_prediction_type: Optional[Literal["object-detection", "classification", "segmentation", "regression"]] = None
...: ml_model_architecture: Optional[str] = None
...:
...: class Config:
...: arbitrary_types_allowed = True
...:
...: # Functions to create, serialize, and deserialize ModelMetadata
...: def create_metadata():
...: input_sig = TensorSignature(name='input_tensor', dtype='float32', shape=(-1, 13, 64, 64))
...: output_sig = TensorSignature(name='output_tensor', dtype='float32', shape=(-1, 10))
...: model_sig = ModelSignature(inputs=[input_sig], outputs=[output_sig])
...: model_artifact = ModelArtifact(path="s3://example/s3/uri/model.pt")
...: class_map = ClassMap(class_to_label_id={
...: 'Annual Crop': 0, 'Forest': 1, 'Herbaceous Vegetation': 2, 'Highway': 3,
...: 'Industrial Buildings': 4, 'Pasture': 5, 'Permanent Crop': 6,
...: 'Residential Buildings': 7, 'River': 8, 'SeaLake': 9
...: })
...: return ModelMetadata(name="eurosat", class_map=class_map, signatures=model_sig, artifact=model_artifact, ml_model_processor_type="cpu")
...:
...: def metadata_json(metadata: ModelMetadata) -> str:
...: return metadata.model_dump_json(indent=2)
...:
...: def model_metadata_json_operations(json_str: str) -> ModelMetadata:
...: return ModelMetadata.model_validate_json(json_str)
...:
...: # Running the functions end-to-end
...: metadata = create_metadata()
...: json_str = metadata_json(metadata)
...: model_metadata = model_metadata_json_operations(json_str)
...:
...: print("Model Metadata Name:", model_metadata.name)
---------------------------------------------------------------------------
NotImplementedError Traceback (most recent call last)
Cell In[1], line 100
98 metadata = create_metadata()
99 json_str = metadata_json(metadata)
--> 100 model_metadata = model_metadata_json_operations(json_str)
102 print("Model Metadata Name:", model_metadata.name)
Cell In[1], line 95, in model_metadata_json_operations(json_str)
94 def model_metadata_json_operations(json_str: str) -> ModelMetadata:
---> 95 return ModelMetadata.model_validate_json(json_str)
File ~/miniforge3/lib/python3.10/site-packages/pydantic/main.py:538, in BaseModel.model_validate_json(cls, json_data, strict, context)
536 # `__tracebackhide__` tells pytest and some other tools to omit this function from tracebacks
537 __tracebackhide__ = True
--> 538 return cls.__pydantic_validator__.validate_json(json_data, strict=strict, context=context)
NotImplementedError: Cannot check isinstance when validating from json, use a JsonOrPython validator instead.
Hi @rbavery,
Apologies for not following up earlier here - feel free to let me know if you're still experiencing this issue in our newest version, and I'd be happy to help!
UlfurOrn
commented
1 month ago Hey 👋 I ran into the same error message while using lambda powertools. E.g.
NotImplementedError: Cannot check issubclass when validating from json, use a JsonOrPython validator instead.
In this case it seems to be caused when a model with a field using the type type[BaseModel] is parsed from JSON.
Heres a minimal example:
import json
from pydantic import BaseModel
class MyWorkingModel(BaseModel):
string: str
submodel: dict[str, Any] | BaseModel
class MyFailingModel(BaseModel):
string: str
submodel: dict[str, Any] | BaseModel | type[BaseModel]
data = {
"string": "string",
"submodel": {"key": "value"},
}
MyWorkingModel.model_validate_json(json.dumps(data)) # works
MyFailingModel.model_validate_json(json.dumps(data)) # fails I've already filed an issue to the Lambda Powertools repo here.
But I wanted to mention this here in case someone else is running into a similar issue (or if this is something that Pydantic wants to improve?).
sydney-runkle
commented
1 month ago Hi,
Thanks for the reproducible example! This is on us - will try to fix here soon.
Initial Checks
Description
I'm attempting to save out an instance of my Pydantic model to json but encountering an error with no info in the traceback.
https://github.com/crim-ca/dlm-extension/blob/41cf8eaa768f41e988ddd693688fc9d795d58cf6/tests/test_schema.py#L11
the following fails with
this error is likely related to https://github.com/pydantic/pydantic/issues/8455 but I couldn't fix this by following this comment's guidance https://github.com/pydantic/pydantic/issues/8455#issuecomment-1873460962
I tried removing the only
isinstancecheck I could find here and it didn't have any effect on the error.Example Code
the package I'm working with can be installed with
pip install stac-model==0.1.1.alpha3then the error can be reproduced with
the model is defined here: https://github.com/crim-ca/dlm-extension/tree/21d1aa9947befc8fb2a46458ef84e9d183d50f48/stac_model
Python, Pydantic & OS Version