Adapt code to make it compatible with pydantic 2.0

Author: Hartorn

.github/workflows/build_backend.yml

@@ -221,8 +221,8 @@ jobs:
       - name: Check Pydantic installed version
         working-directory: python-client 
         run: |
-          pdm run pip freeze | tail -n +1 | grep '^pydantic'
-          pdm run pip freeze | tail -n +1 | grep -q '^pydantic==${{ matrix.pydantic_v1 && '1' || '2' }}\.'
+          pdm run pip freeze | grep '^pydantic'
+          pdm run pip freeze | grep -q '^pydantic==${{ matrix.pydantic_v1 && '1' || '2' }}\.'
 
       - name: Test code
         working-directory: python-client

python-client/giskard/core/validation.py

@@ -22,8 +22,12 @@ def configured_validate_arguments(func):
     # If you check https://docs.pydantic.dev/latest/usage/validation_decorator/#coercion-and-strictness,
     # this explains it will try to convert/coerce type to the type hinting
     # So a string will be "coerced" to an enum element, and so on
-    to_return = functools.wraps(func)(validate_arguments(config={"arbitrary_types_allowed":True})(func))
-    return to_return
+
+    # Add validation wrapper
+    validated_func = validate_arguments(func, config={"arbitrary_types_allowed":True})
+    # Call wraps, to update name, docs, ...
+    validated_func = functools.wraps(func)(validated_func)
+    return func
 
 
 def validate_is_pandasdataframe(df):

python-client/giskard/llm/talk/talk.py

@@ -44,13 +44,17 @@
 Thought: I should look at the features that are required to see what I should gather to predict in the available tools
 {agent_scratchpad}"""
 
+class LenientBaseToolkit(BaseToolkit):
+    """Extended class to allow arbitrary_types_allowed for pydantic compatibility"""
+    class Config:
+        arbitrary_types_allowed = True
+
 
 class ModelSpec(BaseModel):
     """Base class for model spec."""
-
     model: GiskardBaseModel
-    dataset: Optional[Dataset]
-    scan_report: Optional[ScanReport]
+    dataset: Optional[Dataset] = None
+    scan_report: Optional[ScanReport] = None
 
     def _parse_json_inputs(self, json_inputs: str) -> Dict[str, Any]:
         features = json.loads(json_inputs)
@@ -242,7 +246,7 @@ async def _arun(
         return self._run(tool_input)
 
 
-class ModelToolkit(BaseToolkit):
+class ModelToolkit(LenientBaseToolkit):
     """Toolkit for interacting with an ML model."""
 
     spec: ModelSpec

python-client/giskard/ml_worker/testing/registry/decorators.py

@@ -4,7 +4,6 @@
 from typing import Callable, Optional, List, Union, Type, TypeVar
 
 from giskard.core.core import TestFunctionMeta
-from giskard.core.validation import configured_validate_arguments
 from giskard.ml_worker.testing.registry.decorators_utils import make_all_optional_or_suite_input, set_return_type
 from giskard.ml_worker.testing.registry.giskard_test import GiskardTestMethod, GiskardTest
 
@@ -45,4 +44,5 @@ def _wrap_test_method(original):
     make_all_optional_or_suite_input(giskard_test_method)
     set_return_type(giskard_test_method, GiskardTestMethod)
 
-    return configured_validate_arguments(giskard_test_method)()
+    ## why a copy ?
+    return giskard_test_method #configured_validate_arguments(giskard_test_method)#()

python-client/giskard/ml_worker/testing/registry/giskard_test.py

@@ -9,6 +9,7 @@
 import cloudpickle
 
 from giskard.core.core import TestFunctionMeta, SMT
+from giskard.core.validation import configured_validate_arguments
 from giskard.ml_worker.core.savable import Artifact
 from giskard.ml_worker.testing.registry.registry import tests_registry, get_object_uuid
 from giskard.ml_worker.testing.test_result import TestResult
@@ -104,7 +105,7 @@ class GiskardTestMethod(GiskardTest):
     def __init__(self, test_fn: Function) -> None:
         self.params = {}
         self.is_initialized = False
-        self.test_fn = test_fn
+        self.test_fn = configured_validate_arguments(test_fn)
         test_uuid = get_object_uuid(test_fn)
         meta = tests_registry.get_test(test_uuid)
         if meta is None:

python-client/giskard/ml_worker/testing/registry/slicing_function.py

@@ -60,7 +60,7 @@ def __init__(self, func: Optional[SlicingFunctionType], row_level=True, cell_lev
         """
         self.is_initialized = False
         self.params = {}
-        self.func = func
+        self.func = configured_validate_arguments(func)
         self.row_level = row_level
         self.cell_level = cell_level
 
@@ -147,7 +147,7 @@ def inner(func: Union[SlicingFunctionType, Type[SlicingFunction]]) -> SlicingFun
         if inspect.isclass(func) and issubclass(func, SlicingFunction):
             return func
 
-        return _wrap_slicing_function(func, row_level, cell_level)()
+        return _wrap_slicing_function(func, row_level, cell_level)#()
 
     if callable(_fn):
         return functools.wraps(_fn)(inner(_fn))
@@ -166,4 +166,4 @@ def _wrap_slicing_function(original: Callable, row_level: bool, cell_level: bool
     make_all_optional_or_suite_input(slicing_fn)
     set_return_type(slicing_fn, SlicingFunction)
 
-    return configured_validate_arguments(slicing_fn)
+    return slicing_fn #configured_validate_arguments(slicing_fn)

python-client/giskard/ml_worker/testing/registry/transformation_function.py

@@ -32,7 +32,7 @@ def _get_name(cls) -> str:
         return "transformations"
 
     def __init__(self, func: Optional[TransformationFunctionType], row_level=True, cell_level=False):
-        self.func = func
+        self.func = configured_validate_arguments(func)
         self.row_level = row_level
         self.cell_level = cell_level
 
@@ -136,5 +136,5 @@ def _wrap_transformation_function(original: Callable, row_level: bool, cell_leve
 
     make_all_optional_or_suite_input(transformation_fn)
     set_return_type(transformation_fn, TransformationFunction)
-
-    return configured_validate_arguments(transformation_fn)
+    # Copy or not copy ?
+    return transformation_fn #configured_validate_arguments(transformation_fn) #()

python-client/giskard/models/automodel.py

@@ -5,7 +5,7 @@
 
 import pandas as pd
 
-from ..core.core import ModelType
+from ..core.core import ModelType, SupportedModelTypes
 from .base.serialization import CloudpickleSerializableModel
 from .function import PredictionFunctionModel
 
@@ -159,11 +159,11 @@ def __new__(
 
             obj = output_cls(
                 model=model,
-                model_type=model_type,
+                model_type=SupportedModelTypes(model_type) if isinstance(model_type, str) else model_type,
                 data_preprocessing_function=data_preprocessing_function,
                 model_postprocessing_function=model_postprocessing_function,
                 name=name,
-                feature_names=feature_names,
+                feature_names=list(feature_names) if feature_names is not None else None,
                 classification_threshold=classification_threshold,
                 classification_labels=classification_labels,
                 **kwargs,

python-client/giskard/models/base/model.py

@@ -128,7 +128,7 @@ def __init__(
         self.meta = ModelMeta(
             name=name if name is not None else self.__class__.__name__,
             model_type=model_type,
-            feature_names=list(feature_names) if feature_names else None,
+            feature_names=list(feature_names) if feature_names is not None else None,
             classification_labels=np_types_to_native(classification_labels),
             loader_class=self.__class__.__name__,
             loader_module=self.__module__,

python-client/giskard/models/base/model_prediction.py

@@ -1,10 +1,10 @@
 from typing import Any, Optional
 
-import pydantic
+from pydantic import BaseModel, Field
 
 
 # @TODO: Define the fields of this class more rigorously.
-class ModelPredictionResults(pydantic.BaseModel):
+class ModelPredictionResults(BaseModel):
     """Data structure for model predictions.
 
     For regression models, the `prediction` field of the returned `ModelPredictionResults` object will contain the same
@@ -29,8 +29,8 @@ class ModelPredictionResults(pydantic.BaseModel):
         The predicted probabilities for all class labels for each example in the input dataset.
     """
 
-    raw: Any = []
-    prediction: Any = []
-    raw_prediction: Any = []
-    probabilities: Optional[Any]
-    all_predictions: Optional[Any]
+    raw: Any = Field(default_factory=list)
+    prediction: Any = Field(default_factory=list)
+    raw_prediction: Any = Field(default_factory=list)
+    probabilities: Optional[Any] = None
+    all_predictions: Optional[Any] = None