Bump pydantic support to v2

.github/workflows/test-check.yaml

@@ -97,6 +97,6 @@ jobs:
       - name: "Clean sparsezoo directory"
         run: rm -r sparsezoo/
       - name: ⚙️ Install dependencies
-        run: pip install .[dev,haystack]
+        run: pip install .[dev]
       - name: Run integrations tests
         run: make test_integrations

MANIFEST.in

@@ -1,6 +1,5 @@
 include LICENSE
 include utils/artifacts.py
-include src/deepsparse/transformers/haystack/haystack_reqs.txt
 recursive-include src/deepsparse/avx2 *
 recursive-include src/deepsparse/avx512 *
 recursive-include src/deepsparse/neon *

examples/vit_pose/schemas.py

@@ -15,7 +15,7 @@
 from typing import List
 
 import numpy
-from pydantic import BaseModel
+from pydantic import BaseModel, ConfigDict
 
 from deepsparse.pipelines.computer_vision import ComputerVisionSchema
 
@@ -32,6 +32,4 @@ class VitPoseInput(ComputerVisionSchema):
 
 class VitPoseOutput(BaseModel):
     out: List[numpy.ndarray]
-
-    class Config:
-        arbitrary_types_allowed = True
+    model_config = ConfigDict(arbitrary_types_allowed=True)

integrations/test_placeholder.py

@@ -0,0 +1,14 @@
+
+def test_placeholder():
+    """
+    Needed to make the test suite run and not throw
+    an error about no tests being found when
+    `make test_integrations` is used.
+
+    The error would look like this:
+        make: *** [Makefile:61: test_integrations] Error 5
+
+    More information can be found here:
+        https://github.com/pytest-dev/pytest/issues/2393
+    """
+    pass

setup.py

@@ -77,17 +77,10 @@
 ]
 
 
-def _parse_requirements_file(file_path):
-    with open(file_path, "r") as requirements_file:
-        lines = requirements_file.read().splitlines()
-
-    return [line for line in lines if len(line) > 0 and line[0] != "#"]
-
-
 _deps = [
     "numpy>=1.16.3",
     "onnx>=1.5.0,<1.15.0",
-    "pydantic>=1.8.2,<2.0.0",
+    "pydantic>=2.0.0,<2.8.0",
     "requests>=2.0.0",
     "tqdm>=4.0.0",
     "protobuf>=3.12.2",
@@ -122,7 +115,7 @@ def _parse_requirements_file(file_path):
 ]
 _server_deps = [
     "uvicorn>=0.15.0",
-    "fastapi>=0.70.0,<0.87.0",
+    "fastapi>=0.100.0,<0.111",
     "requests>=2.26.0",
     "python-multipart>=0.0.5",
     "prometheus-client>=0.14.1",
@@ -153,17 +146,6 @@ def _parse_requirements_file(file_path):
 ]
 _sentence_transformers_integration_deps = ["optimum-deepsparse"] + _torch_deps
 
-# haystack dependencies are installed from a requirements file to avoid
-# conflicting versions with NM's deepsparse/transformers
-_haystack_requirements_file_path = os.path.join(
-    os.path.dirname(os.path.realpath(__file__)),
-    "src",
-    "deepsparse",
-    "transformers",
-    "haystack",
-    "haystack_reqs.txt",
-)
-_haystack_integration_deps = _parse_requirements_file(_haystack_requirements_file_path)
 _clip_deps = [
     "open_clip_torch==2.20.0",
     "transformers<4.40",
@@ -270,7 +252,6 @@ def _setup_extras() -> Dict:
         "image_classification": _computer_vision_deps,
         "yolo": _computer_vision_deps,
         "yolov5": _computer_vision_deps,
-        "haystack": _haystack_integration_deps,
         "openpifpaf": _openpifpaf_integration_deps,
         "yolov8": _yolov8_integration_deps,
         "transformers": _transformers_integration_deps,

src/deepsparse/benchmark/data_creation.py

@@ -17,7 +17,7 @@
 import random
 import string
 from os import path
-from typing import Dict, List, Tuple
+from typing import Dict, List, Tuple, get_args
 
 import numpy
 
@@ -58,15 +58,11 @@ def get_input_schema_type(pipeline: Pipeline) -> str:
     if SchemaType.TEXT_SEQ in input_schema_requirements:
         if input_schema_fields.get(SchemaType.TEXT_SEQ).alias == SchemaType.TEXT_PROMPT:
             return SchemaType.TEXT_PROMPT
-        sequence_types = [
-            f.outer_type_ for f in input_schema_fields[SchemaType.TEXT_SEQ].sub_fields
-        ]
+        sequence_types = get_args(input_schema_fields[SchemaType.TEXT_SEQ].annotation)
         if List[str] in sequence_types:
             return SchemaType.TEXT_SEQ
     elif SchemaType.TEXT_INPUT in input_schema_requirements:
-        sequence_types = [
-            f.outer_type_ for f in input_schema_fields[SchemaType.TEXT_INPUT].sub_fields
-        ]
+        sequence_types = get_args(input_schema_fields[SchemaType.TEXT_INPUT].annotation)
         if List[str] in sequence_types:
             return SchemaType.TEXT_INPUT
     elif SchemaType.QUESTION in input_schema_requirements:

src/deepsparse/clip/decoder_pipeline.py

@@ -30,10 +30,10 @@ class CLIPDecoderInput(BaseModel):
     """
 
     text_embeddings: Any = Field(
-        description="np.array of text emebddings from the " "text branch"
+        None, description="np.array of text emebddings from the " "text branch"
     )
     image_embeddings: Any = Field(
-        description="np.array of image embeddings from the " "visual branch"
+        None, description="np.array of image embeddings from the " "visual branch"
     )
 
 

src/deepsparse/evaluation/results.py

@@ -36,15 +36,15 @@ class Metric(BaseModel):
 
 
 class Dataset(BaseModel):
-    type: Optional[str] = Field(description="Type of dataset")
+    type: Optional[str] = Field(None, description="Type of dataset")
     name: str = Field(description="Name of the dataset")
-    config: Any = Field(description="Configuration for the dataset")
-    split: Optional[str] = Field(description="Split of the dataset")
+    config: Any = Field(None, description="Configuration for the dataset")
+    split: Optional[str] = Field(None, description="Split of the dataset")
 
 
 class EvalSample(BaseModel):
-    input: Any = Field(description="Sample input to the model")
-    output: Any = Field(description="Sample output from the model")
+    input: Any = Field(None, description="Sample input to the model")
+    output: Any = Field(None, description="Sample output from the model")
 
 
 class Evaluation(BaseModel):
@@ -55,7 +55,7 @@ class Evaluation(BaseModel):
     dataset: Dataset = Field(description="Dataset that the evaluation was performed on")
     metrics: List[Metric] = Field(description="List of metrics for the evaluation")
     samples: Optional[List[EvalSample]] = Field(
-        description="List of samples for the evaluation"
+        None, description="List of samples for the evaluation"
     )
 
 
@@ -64,8 +64,9 @@ class Result(BaseModel):
         description="Evaluation result represented in the unified, structured format"
     )
     raw: Any = Field(
+        None,
         description="Evaluation result represented in the raw format "
-        "(characteristic for the specific evaluation integration)"
+        "(characteristic for the specific evaluation integration)",
     )
 
 
@@ -97,7 +98,7 @@ def _save_to_json(result: Result, save_path: str):
 
 
 def _save_to_yaml(result: Result, save_path: str):
-    _save(yaml.dump(result.dict()), save_path, expected_ext=".yaml")
+    _save(yaml.dump(result.model_dump()), save_path, expected_ext=".yaml")
 
 
 def _save(data: str, save_path: str, expected_ext: str):

src/deepsparse/legacy/loggers/config.py

@@ -14,7 +14,7 @@
 
 from typing import Any, Dict, List, Optional
 
-from pydantic import BaseModel, Field, validator
+from pydantic import BaseModel, Field, field_validator
 
 
 """
@@ -57,7 +57,8 @@ class MetricFunctionConfig(BaseModel):
         "the subset of loggers (specified here by a list of their names).",
     )
 
-    @validator("frequency")
+    @field_validator("frequency")
+    @classmethod
     def non_zero_frequency(cls, frequency: int) -> int:
         if frequency <= 0:
             raise ValueError(

src/deepsparse/legacy/tasks.py

@@ -206,7 +206,10 @@ def check_register_task(
         elif cls.is_haystack(task):
             # trigger haystack pipeline as well as transformers pipelines to
             # register with Pipeline.register
-            import deepsparse.transformers.haystack  # noqa: F401
+            raise DeprecationWarning(
+                "Haystack support with deepsparse has been deprecated, "
+                "kindly use deepsparse-nightly==1.8.20240404 or older"
+            )
 
         elif cls.is_embedding_extraction(task):
             # trigger embedding_extraction pipelines to register with

src/deepsparse/loggers/config.py

@@ -15,12 +15,11 @@
 from typing import Dict, List, Optional
 
 import yaml
-from pydantic import BaseModel, Extra, Field, validator
+from pydantic import BaseModel, ConfigDict, Field, validator
 
 
 class LoggerConfig(BaseModel):
-    class Config:
-        extra = Extra.allow
+    model_config = ConfigDict(extra="allow")
 
     name: str = Field(
         default="PythonLogger",

src/deepsparse/loggers/logger_manager.py

@@ -65,7 +65,7 @@ class LoggerManager(AsyncExecutor, LoggerFactory):
     """
 
     def __init__(self, config: str = ""):
-        self.config = LoggingConfig.from_config(config).dict()
+        self.config = LoggingConfig.from_config(config).model_dump()
         super().__init__(config=self.config)
 
     def log(

src/deepsparse/loggers/root_logger.py

@@ -38,7 +38,7 @@ class RootLogger(FrequencyFilter):
     its own FrequencyFilter
 
     :param config: config with respect to
-     the log_type (LoggerConfig().dict().get(log_type))
+     the log_type (LoggerConfig().model_dump().get(log_type))
     :param leaf_logger: leaf logger singleton shared among other RootLogger
 
     """

src/deepsparse/open_pif_paf/schemas.py

@@ -16,7 +16,7 @@
 
 import numpy
 from PIL import Image
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, ConfigDict, Field
 
 from deepsparse.pipelines.computer_vision import ComputerVisionSchema
 
@@ -76,8 +76,7 @@ def from_files(
         input_schema = cls(*args, images=files_numpy, **kwargs)
         return input_schema
 
-    class Config:
-        arbitrary_types_allowed = True
+    model_config = ConfigDict(arbitrary_types_allowed=True)
 
 
 class OpenPifPafOutput(BaseModel):
@@ -105,6 +104,4 @@ class OpenPifPafOutput(BaseModel):
         "For every prediction, it is a list of tuples of body "
         "part indices. "
     )
-
-    class Config:
-        arbitrary_types_allowed = True
+    model_config = ConfigDict(arbitrary_types_allowed=True)