From 0f4eced5d78b8da09c0db2db8e8cf4ced1a68f0c Mon Sep 17 00:00:00 2001
From: BIGAUD Nathan <nathan.bigaud@inria.fr>
Date: Tue, 7 Mar 2023 12:15:10 +0100
Subject: [PATCH] Applying corrections after MR review: * Re-introducing
deprecated `DataField` subclasses with a warning * Other minor changes, to
`sklearn` and `tf` models, and `DataTypeField`
---
declearn/data_info/__init__.py | 6 +-
declearn/data_info/_fields.py | 143 ++++++++++++++++++++--------
declearn/dataset/_base.py | 4 +-
declearn/dataset/_inmemory.py | 2 +-
declearn/model/sklearn/_sgd.py | 11 ++-
declearn/model/tensorflow/_model.py | 25 ++---
test/functional/test_regression.py | 2 +-
7 files changed, 133 insertions(+), 60 deletions(-)
diff --git a/declearn/data_info/__init__.py b/declearn/data_info/__init__.py
index 7dc304e9..4b4ede5f 100644
--- a/declearn/data_info/__init__.py
+++ b/declearn/data_info/__init__.py
@@ -57,7 +57,9 @@ from ._base import (
)
from ._fields import (
ClassesField,
- SingleInputShapeField,
+ DataTypeField,
+ InputShapeField,
+ NbFeaturesField,
NbSamplesField,
- DataTypeField
+ SingleInputShapeField,
)
diff --git a/declearn/data_info/_fields.py b/declearn/data_info/_fields.py
index 2f27461b..dd0b0bcb 100644
--- a/declearn/data_info/_fields.py
+++ b/declearn/data_info/_fields.py
@@ -17,7 +17,8 @@
"""DataInfoField subclasses specifying common 'data_info' metadata fields."""
-from typing import Any, ClassVar, Optional, Set, Tuple, Type
+from typing import Any, ClassVar, List, Optional, Set, Tuple, Type
+from warnings import warn
import numpy as np
@@ -27,6 +28,8 @@ __all__ = [
"ClassesField",
"SingleInputShapeField",
"NbSamplesField",
+ "NbFeaturesField",
+ "InputShapeField",
]
@@ -91,39 +94,7 @@ class SingleInputShapeField(DataInfoField):
f"Cannot combine '{cls.field}': inputs don't have the same"
"shape"
)
- return unique_shapes[0] # type: ignore
-
-
-# @register_data_info_field
-# class NbFeaturesField(DataInfoField):
-# """Specifications for 'n_features' data_info field."""
-
-# field: ClassVar[str] = "n_features"
-# types: ClassVar[Tuple[Type, ...]] = (int,)
-# doc: ClassVar[str] = "Number of input features, checked to be equal."
-
-# @classmethod
-# def is_valid(
-# cls,
-# value: Any,
-# ) -> bool:
-# return isinstance(value, int) and (value > 0)
-
-# @classmethod
-# def combine(
-# cls,
-# *values: Any,
-# ) -> int:
-# unique = list(set(values))
-# if len(unique) != 1:
-# raise ValueError(
-# f"Cannot combine '{cls.field}': non-unique inputs."
-# )
-# if not cls.is_valid(unique[0]):
-# raise ValueError(
-# f"Cannot combine '{cls.field}': invalid unique value."
-# )
-# return unique[0]
+ return unique_shapes[0] # type: ignore
@register_data_info_field
@@ -165,16 +136,108 @@ class DataTypeField(DataInfoField):
) -> bool:
out = isinstance(value, str)
if out:
- # CHECK
try:
np.dtype(value)
- except TypeError as exp:
- raise TypeError(
- "The received string could not be parsed"
- "to a valid array dtype"
- ) from exp
+ except TypeError:
+ out = False
return out
+ @classmethod
+ def combine(
+ cls,
+ *values: Any,
+ ) -> int:
+ super().combine(*values)
+ unique = list(set(values))
+ if len(unique) != 1:
+ raise ValueError(
+ f"Cannot combine '{cls.field}': non-unique inputs."
+ )
+ if not cls.is_valid(unique[0]):
+ raise ValueError(
+ f"Cannot combine '{cls.field}': invalid unique value."
+ )
+ return unique[0]
+
+
+@register_data_info_field
+class InputShapeField(DataInfoField):
+ """Deprecated specifications for 'input_shape' data_info field."""
+
+ field: ClassVar[str] = "input_shape"
+ types: ClassVar[Tuple[Type, ...]] = (tuple, list)
+ doc: ClassVar[str] = "Input features' batched shape, checked to be equal."
+
+ def __init__(self) -> None:
+ warn(
+ f"{self.__class__.__name__} will be deprecated.",
+ DeprecationWarning,
+ stacklevel=2,
+ )
+
+ @classmethod
+ def is_valid(
+ cls,
+ value: Any,
+ ) -> bool:
+ return (
+ isinstance(value, cls.types)
+ and (len(value) >= 2)
+ and all(isinstance(val, int) or (val is None) for val in value)
+ )
+
+ @classmethod
+ def combine(
+ cls,
+ *values: Any,
+ ) -> List[Optional[int]]:
+ # Type check each and every input shape.
+ super().combine(*values)
+ # Check that all shapes are of same length.
+ unique = list({len(shp) for shp in values})
+ if len(unique) != 1:
+ raise ValueError(
+ f"Cannot combine '{cls.field}': inputs have various lengths."
+ )
+ # Fill-in the unified shape: except all-None or (None or unique) value.
+ # Note: batching dimension is set to None by default (no check).
+ shape = [None] * unique[0] # type: List[Optional[int]]
+ for i in range(1, unique[0]):
+ val = [shp[i] for shp in values if shp[i] is not None]
+ if not val: # all None
+ shape[i] = None
+ elif len(set(val)) > 1:
+ raise ValueError(
+ f"Cannot combine '{cls.field}': provided shapes differ."
+ )
+ else:
+ shape[i] = val[0]
+ # Return the combined shape.
+ return shape
+
+
+@register_data_info_field
+class NbFeaturesField(DataInfoField):
+ """Deprecated specifications for 'n_features' data_info field."""
+
+ field: ClassVar[str] = "n_features"
+ types: ClassVar[Tuple[Type, ...]] = (int,)
+ doc: ClassVar[str] = "Number of input features, checked to be equal."
+
+ def __init__(self) -> None:
+ warn(
+ f"{self.__class__.__name__} will be deprecated.",
+ DeprecationWarning,
+ stacklevel=2,
+ )
+
+ @classmethod
+ def is_valid(
+ cls,
+ value: Any,
+ ) -> bool:
+ return isinstance(value, int) and (value > 0)
+
@classmethod
def combine(
cls,
diff --git a/declearn/dataset/_base.py b/declearn/dataset/_base.py
index 4e201fca..a7e9c167 100644
--- a/declearn/dataset/_base.py
+++ b/declearn/dataset/_base.py
@@ -19,7 +19,7 @@
from abc import ABCMeta, abstractmethod
from dataclasses import dataclass
-from typing import Any, ClassVar, Iterator, Optional, Set, Tuple, Union, List
+from typing import Any, ClassVar, Iterator, List, Optional, Set, Tuple, Union
from typing_extensions import Self # future: import from typing (py >=3.11)
@@ -38,7 +38,7 @@ class DataSpecs:
"""Dataclass to wrap a dataset's metadata."""
n_samples: int
- single_input_shape: Union[Tuple[int],List[int]]
+ single_input_shape: Union[Tuple[int], List[int]]
classes: Optional[Set[Any]] = None
data_type: Optional[str] = None
diff --git a/declearn/dataset/_inmemory.py b/declearn/dataset/_inmemory.py
index 51d33d2a..6efd2714 100644
--- a/declearn/dataset/_inmemory.py
+++ b/declearn/dataset/_inmemory.py
@@ -440,7 +440,7 @@ class InMemoryDataset(Dataset):
"""Return a DataSpecs object describing this dataset."""
return DataSpecs(
n_samples=self.feats.shape[0],
- single_input_shape=self.feats.shape[1:], # type: ignore
+ single_input_shape=self.feats.shape[1:], # type: ignore
classes=self.classes,
data_type=self.data_type,
)
diff --git a/declearn/model/sklearn/_sgd.py b/declearn/model/sklearn/_sgd.py
index d4d48a4c..cc5ee6e3 100644
--- a/declearn/model/sklearn/_sgd.py
+++ b/declearn/model/sklearn/_sgd.py
@@ -136,9 +136,13 @@ class SklearnSGDModel(Model):
self._model.classes_ = np.array(list(data_info["classes"]))
n_classes = len(self._model.classes_)
dim = n_classes if (n_classes > 2) else 1
- self._model.coef_ = np.zeros(
- (dim, *data_info["single_input_shape"])
+ if len(data_info["single_input_shape"]) != 1:
+ raise ValueError(
+ "SklearnSGDModel currently only supports"
+ "flat, one dimensional features"
)
+ feat = data_info["single_input_shape"][0]
+ self._model.coef_ = np.zeros((dim, feat))
self._model.intercept_ = np.zeros((dim,))
# SGDRegressor case.
else:
@@ -380,7 +384,7 @@ class SklearnSGDModel(Model):
) -> Callable[[np.ndarray, np.ndarray], np.ndarray]:
"""Return a function to compute point-wise loss for a given batch."""
# fmt: off
- # Gather or instantiate a loss function from the wrapped model's specs.
+ # Gather / instantiate a loss function from the wrapped model's specs.
if hasattr(self._model, "loss_function_"):
loss_smp = self._model.loss_function_.py_loss
else:
@@ -399,6 +403,7 @@ class SklearnSGDModel(Model):
else:
loss_fn = loss_1d
return loss_fn
+ # fmt: on
def update_device_policy(
self,
diff --git a/declearn/model/tensorflow/_model.py b/declearn/model/tensorflow/_model.py
index 452608cb..10863a24 100644
--- a/declearn/model/tensorflow/_model.py
+++ b/declearn/model/tensorflow/_model.py
@@ -29,9 +29,11 @@ from declearn.data_info import aggregate_data_info
from declearn.model._utils import raise_on_stringsets_mismatch
from declearn.model.api import Model
from declearn.model.tensorflow._vector import TensorflowVector
-from declearn.model.tensorflow.utils import (build_keras_loss,
- move_layer_to_device,
- select_device)
+from declearn.model.tensorflow.utils import (
+ build_keras_loss,
+ move_layer_to_device,
+ select_device,
+)
from declearn.typing import Batch
from declearn.utils import DevicePolicy, get_device_policy, register_type
@@ -60,7 +62,7 @@ class TensorflowModel(Model):
* Note that if the global device-placement policy is updated, this will
only be propagated to existing instances by manually calling their
`update_device_policy` method.
- * You may consult the device policy currently enforced by a TensorflowModel
+ * You may consult the device policy enforced by a TensorflowModel
instance by accessing its `device_policy` property.
"""
@@ -129,20 +131,20 @@ class TensorflowModel(Model):
def required_data_info(
self,
) -> Set[str]:
- return set() if self._model.built else {"n_samples",
- "single_input_shape"}
+ return (
+ set() if self._model.built else {"n_samples", "single_input_shape"}
+ )
def initialize(
self,
data_info: Dict[str, Any],
) -> None:
if not self._model.built:
- data_info = aggregate_data_info([data_info],
- self.required_data_info)
+ data_info = aggregate_data_info(
+ [data_info], self.required_data_info
+ )
with tf.device(self._device):
- self._model.build(
- (data_info['n_samples'], *data_info['single_input_shape'])
- )
+ self._model.build((None, *data_info["single_input_shape"]))
def get_config(
self,
@@ -256,6 +258,7 @@ class TensorflowModel(Model):
return tf.convert_to_tensor(data)
# Apply it to the the batched elements.
return tf.nest.map_structure(convert, batch)
+ # fmt: on
@tf.function # optimize tensorflow runtime
def _compute_batch_gradients(
diff --git a/test/functional/test_regression.py b/test/functional/test_regression.py
index 70a52924..d08b5f16 100644
--- a/test/functional/test_regression.py
+++ b/test/functional/test_regression.py
@@ -285,7 +285,7 @@ def test_declearn_baseline(
d_train = InMemoryDataset(train[0], train[1])
# Set up a declearn model and a vanilla SGD optimizer.
model = get_model("numpy")
- model.initialize({"n_features": d_train.data.shape[1]})
+ model.initialize({"single_input_shape": (d_train.data.shape[1],)})
opt = Optimizer(lrate=lrate, regularizers=[("lasso", {"alpha": 0.1})])
# Iteratively train the model, evaluating it after each epoch.
for _ in range(rounds):
--
GitLab