Merge pull request #78 from online-ml/rename_deep_river

.github/workflows/unit-tests.yml

@@ -43,6 +43,6 @@ jobs:
         run: python -c "from river import datasets; datasets.CreditCard().download(); datasets.Elec2().download(); datasets.Keystroke().download()"
 
       - name: pytest
-        run: pytest --cov=river_torch -m "not datasets"
+        run: pytest --cov=deep_river -m "not datasets"
       - name: Upload coverage reports to Codecov with GitHub Action
         uses: codecov/codecov-action@v2

.pre-commit-config.yaml

@@ -1,4 +1,4 @@
-files: river_torch
+files: deep_river
 repos:
   - repo: https://github.com/pre-commit/pre-commit-hooks
     rev: v4.2.0

README.md

@@ -1,46 +1,46 @@
 <p align="center">
-  <img height="150px" src="https://raw.githubusercontent.com/online-ml/river-torch/master/docs/img/logo.png" alt="incremental dl logo">
+  <img height="150px" src="https://raw.githubusercontent.com/online-ml/deep-river/master/docs/img/logo.png" alt="incremental dl logo">
 </p>
 <p align="center">
-    <img alt="PyPI" src="https://img.shields.io/pypi/v/river-torch">
-    <a href="https://codecov.io/gh/online-ml/river-torch" > 
-        <img src="https://codecov.io/gh/online-ml/river-torch/branch/master/graph/badge.svg?token=ZKUIISZAYA"/> 
+    <img alt="PyPI" src="https://img.shields.io/pypi/v/deep-river">
+    <a href="https://codecov.io/gh/online-ml/deep-river" > 
+        <img src="https://codecov.io/gh/online-ml/deep-river/branch/master/graph/badge.svg?token=ZKUIISZAYA"/> 
     </a>
-    <img alt="PyPI - Downloads" src="https://img.shields.io/pypi/dw/river-torch">
-    <img alt="GitHub" src="https://img.shields.io/github/license/online-ml/river-torch">
+    <img alt="PyPI - Downloads" src="https://img.shields.io/pypi/dw/deep-river">
+    <img alt="GitHub" src="https://img.shields.io/github/license/online-ml/deep-river">
 </p>
 <p align="center">
-    river-torch is a Python library for online deep learning.
-    River-torch's ambition is to enable <a href="https://www.wikiwand.com/en/Online_machine_learning">online machine learning</a> for neural networks.
+    deep-river is a Python library for online deep learning.
+    deep-river's ambition is to enable <a href="https://www.wikiwand.com/en/Online_machine_learning">online machine learning</a> for neural networks.
     It combines the <a href="https://www.riverml.xyz">river</a> API with the capabilities of designing neural networks based on <a href="https://pytorch.org">PyTorch</a>.
 </p>
 
 ## 💈 Installation
 
 ```shell
-pip install river-torch
+pip install deep-river
 ```
 or
 ```shell
-pip install "river[torch]"
+pip install "river[deep]"
 ```
 You can install the latest development version from GitHub as so:
 
 ```shell
-pip install https://github.com/online-ml/river-torch/archive/refs/heads/master.zip
+pip install https://github.com/online-ml/deep-river/archive/refs/heads/master.zip
 ```
 
 ## 🍫 Quickstart
 
 We build the development of neural networks on top of the <a href="https://www.riverml.xyz">river API</a> and refer to the rivers design principles.
 The following example creates a simple MLP architecture based on PyTorch and incrementally predicts and trains on the website phishing dataset.
-For further examples check out the <a href="https://online-ml.github.io/river-torch">Documentation</a>.
+For further examples check out the <a href="https://online-ml.github.io/deep-river">Documentation</a>.
 
 ### Classification
 
 ```python
 >>> from river import metrics, datasets, preprocessing, compose
->>> from river_torch import classification
+>>> from deep_river import classification
 >>> from torch import nn
 >>> from torch import optim
 >>> from torch import manual_seed
@@ -72,7 +72,7 @@ For further examples check out the <a href="https://online-ml.github.io/river-to
 >>> for x, y in dataset:
 ...     y_pred = model_pipeline.predict_one(x)  # make a prediction
 ...     metric = metric.update(y, y_pred)  # update the metric
-...     model_pipeline = model_pipeline.learn_one(x,y)  # make the model learn
+...     model_pipeline = model_pipeline.learn_one(x, y)  # make the model learn
 >>> print(f"Accuracy: {metric.get():.4f}")
 Accuracy: 0.6728
 
@@ -81,7 +81,7 @@ Accuracy: 0.6728
 ### Anomaly Detection
 
 ```python
->>> from river_torch.anomaly import Autoencoder
+>>> from deep_river.anomaly import Autoencoder
 >>> from river import metrics
 >>> from river.datasets import CreditCard
 >>> from torch import nn
@@ -111,9 +111,9 @@ Accuracy: 0.6728
 >>> model = Pipeline(scaler, ae)
 
 >>> for x, y in dataset:
-...    score = model.score_one(x)
-...    model = model.learn_one(x=x)
-...    metric = metric.update(y, score)
+...     score = model.score_one(x)
+...     model = model.learn_one(x=x)
+...     metric = metric.update(y, score)
 ...
 >>> print(f"ROCAUC: {metric.get():.4f}")
 ROCAUC: 0.7447

river_torch/anomaly/ae.py → deep_river/anomaly/ae.py

@@ -6,9 +6,9 @@
 from river.anomaly.base import AnomalyDetector
 from torch import nn
 
-from river_torch.base import DeepEstimator
-from river_torch.utils import dict2tensor
-from river_torch.utils.tensor_conversion import df2tensor
+from deep_river.base import DeepEstimator
+from deep_river.utils import dict2tensor
+from deep_river.utils.tensor_conversion import df2tensor
 
 
 class _TestAutoencoder(torch.nn.Module):
@@ -58,7 +58,7 @@ class Autoencoder(DeepEstimator, AnomalyDetector):
 
     Examples
     --------
-    >>> from river_torch.anomaly import Autoencoder
+    >>> from deep_river.anomaly import Autoencoder
     >>> from river import metrics
     >>> from river.datasets import CreditCard
     >>> from torch import nn

river_torch/anomaly/probability_weighted_ae.py → deep_river/anomaly/probability_weighted_ae.py

@@ -6,8 +6,8 @@
 from river import stats, utils
 from scipy.special import ndtr
 
-from river_torch.anomaly import ae
-from river_torch.utils import dict2tensor
+from deep_river.anomaly import ae
+from deep_river.utils import dict2tensor
 
 
 class ProbabilityWeightedAutoencoder(ae.Autoencoder):
@@ -51,7 +51,7 @@ class ProbabilityWeightedAutoencoder(ae.Autoencoder):
 
         Examples
     --------
-    >>> from river_torch.anomaly import ProbabilityWeightedAutoencoder
+    >>> from deep_river.anomaly import ProbabilityWeightedAutoencoder
     >>> from river import metrics
     >>> from river.datasets import CreditCard
     >>> from torch import nn, manual_seed

river_torch/anomaly/rolling_ae.py → deep_river/anomaly/rolling_ae.py

@@ -6,8 +6,8 @@
 from river import anomaly
 from torch import nn
 
-from river_torch.base import RollingDeepEstimator
-from river_torch.utils.tensor_conversion import deque2rolling_tensor
+from deep_river.base import RollingDeepEstimator
+from deep_river.utils.tensor_conversion import deque2rolling_tensor
 
 
 class _TestLSTMAutoencoder(nn.Module):

river_torch/base.py → deep_river/base.py

@@ -6,7 +6,7 @@
 import torch
 from river import base
 
-from river_torch.utils import get_loss_fn, get_optim_fn
+from deep_river.utils import get_loss_fn, get_optim_fn
 
 try:
     from graphviz import Digraph

deep_river/classification/__init__.py

@@ -0,0 +1,7 @@
+from deep_river.classification.classifier import Classifier
+from deep_river.classification.rolling_classifier import RollingClassifier
+
+__all__ = [
+    "Classifier",
+    "RollingClassifier",
+]

river_torch/classification/classifier.py → deep_river/classification/classifier.py

@@ -10,9 +10,9 @@
 from torch import nn
 from torch.utils.hooks import RemovableHandle
 
-from river_torch.base import DeepEstimator
-from river_torch.utils.hooks import ForwardOrderTracker, apply_hooks
-from river_torch.utils.tensor_conversion import (
+from deep_river.base import DeepEstimator
+from deep_river.utils.hooks import ForwardOrderTracker, apply_hooks
+from deep_river.utils.tensor_conversion import (
     df2tensor,
     dict2tensor,
     labels2onehot,
@@ -35,7 +35,7 @@ def forward(self, X, **kwargs):
         return X
 
 
-class Classifier(DeepEstimator, base.Classifier):
+class Classifier(DeepEstimator, base.MiniBatchClassifier):
     """
         Wrapper for PyTorch classification models that automatically handles
         increases in the number of classes by adding output neurons in case
@@ -83,7 +83,7 @@ class Classifier(DeepEstimator, base.Classifier):
         Examples
         --------
     >>> from river import metrics, preprocessing, compose, datasets
-    >>> from river_torch import classification
+    >>> from deep_river import classification
     >>> from torch import nn
     >>> from torch import manual_seed
 
@@ -258,7 +258,7 @@ def predict_proba_one(self, x: dict) -> Dict[ClfTarget, float]:
             y_pred = self.module(x_t)
         return output2proba(
             y_pred, self.observed_classes, self.output_is_logit
-        )
+        )[0]
 
     def learn_many(self, X: pd.DataFrame, y: pd.Series) -> "Classifier":
         """
@@ -286,7 +286,7 @@ def learn_many(self, X: pd.DataFrame, y: pd.Series) -> "Classifier":
         if self.is_class_incremental:
             self._adapt_output_dim()
 
-        return self._learn(x=X, y=y.tolist())
+        return self._learn(x=X, y=y)
 
     def predict_proba_many(self, X: pd.DataFrame) -> pd.DataFrame:
         """
@@ -309,7 +309,7 @@ def predict_proba_many(self, X: pd.DataFrame) -> pd.DataFrame:
         self.module.eval()
         with torch.inference_mode():
             y_preds = self.module(X_t)
-        return pd.Dataframe(output2proba(y_preds, self.observed_classes))
+        return pd.DataFrame(output2proba(y_preds, self.observed_classes))
 
     def _adapt_output_dim(self):
         out_features_target = (

river_torch/classification/rolling_classifier.py → deep_river/classification/rolling_classifier.py

@@ -6,9 +6,9 @@
 from river.base.typing import ClfTarget
 from torch import nn
 
-from river_torch.base import RollingDeepEstimator
-from river_torch.classification import Classifier
-from river_torch.utils.tensor_conversion import (
+from deep_river.base import RollingDeepEstimator
+from deep_river.classification import Classifier
+from deep_river.utils.tensor_conversion import (
     deque2rolling_tensor,
     output2proba,
 )
@@ -80,7 +80,7 @@ class RollingClassifier(Classifier, RollingDeepEstimator):
 
     Examples
     --------
-    >>> from river_torch.classification import RollingClassifier
+    >>> from deep_river.classification import RollingClassifier
     >>> from river import metrics, datasets, compose, preprocessing
     >>> import torch
 
@@ -266,7 +266,7 @@ def predict_proba_one(self, x: dict) -> Dict[ClfTarget, float]:
         if self.append_predict:
             self._x_window.append(list(x.values()))
 
-        return proba
+        return proba[0]
 
     def learn_many(self, X: pd.DataFrame, y: pd.Series) -> "RollingClassifier":
         """
@@ -333,7 +333,7 @@ def predict_proba_many(self, X: pd.DataFrame) -> pd.DataFrame:
             probas = [default_proba] * len(X)
         return pd.DataFrame(probas)
 
-    def _get_default_proba(self):
+    def _get_default_proba(self) -> List[Dict[ClfTarget, float]]:
         if len(self.observed_classes) > 0:
             mean_proba = (
                 1 / len(self.observed_classes)
@@ -343,7 +343,7 @@ def _get_default_proba(self):
             proba = {c: mean_proba for c in self.observed_classes}
         else:
             proba = {c: 1.0 for c in self.observed_classes}
-        return proba
+        return [proba] if isinstance(proba, dict) else proba
 
     def _adapt_output_dim(self):
         out_features_target = (

deep_river/regression/__init__.py

@@ -0,0 +1,7 @@
+from deep_river.regression.regressor import Regressor
+from deep_river.regression.rolling_regressor import RollingRegressor
+
+__all__ = [
+    "Regressor",
+    "RollingRegressor",
+]

river_torch/regression/regressor.py → deep_river/regression/regressor.py

@@ -5,8 +5,8 @@
 from river import base
 from river.base.typing import RegTarget
 
-from river_torch.base import DeepEstimator
-from river_torch.utils.tensor_conversion import (
+from deep_river.base import DeepEstimator
+from deep_river.utils.tensor_conversion import (
     df2tensor,
     dict2tensor,
     float2tensor,
@@ -32,7 +32,7 @@ def forward(self, X, **kwargs):
         return X
 
 
-class Regressor(DeepEstimator, base.Regressor):
+class Regressor(DeepEstimator, base.MiniBatchRegressor):
     """
     Wrapper for PyTorch regression models that enables
     compatibility with River.
@@ -181,7 +181,7 @@ def predict_one(self, x: dict) -> RegTarget:
             y_pred = self.module(x_t).item()
         return y_pred
 
-    def learn_many(self, X: pd.DataFrame, y: List) -> "Regressor":
+    def learn_many(self, X: pd.DataFrame, y: pd.Series) -> "Regressor":
         """
         Performs one step of training with a batch of examples.
 
@@ -201,7 +201,9 @@ def learn_many(self, X: pd.DataFrame, y: List) -> "Regressor":
             self.kwargs["n_features"] = len(X.columns)
             self.initialize_module(**self.kwargs)
         X_t = df2tensor(X, device=self.device)
-        y_t = torch.tensor(y, device=self.device, dtype=torch.float32)
+        y_t = torch.tensor(
+            y, device=self.device, dtype=torch.float32
+        ).unsqueeze(1)
         self._learn(X_t, y_t)
         return self
 
@@ -224,7 +226,6 @@ def predict_many(self, X: pd.DataFrame) -> List:
             self.initialize_module(**self.kwargs)
 
         X = df2tensor(X, device=self.device)
-        self.module.eval()
         with torch.inference_mode():
-            y_preds = self.module(X).detach().tolist()
+            y_preds = self.module(X).detach().squeeze().tolist()
         return y_preds

river_torch/regression/rolling_regressor.py → deep_river/regression/rolling_regressor.py

@@ -4,9 +4,9 @@
 import torch
 from river.base.typing import RegTarget
 
-from river_torch.base import RollingDeepEstimator
-from river_torch.regression import Regressor
-from river_torch.utils.tensor_conversion import (
+from deep_river.base import RollingDeepEstimator
+from deep_river.regression import Regressor
+from deep_river.utils.tensor_conversion import (
     deque2rolling_tensor,
     float2tensor,
 )

river_torch/utils/tensor_conversion.py → deep_river/utils/tensor_conversion.py

@@ -1,4 +1,4 @@
-from typing import Deque, Dict, Optional, Union
+from typing import Deque, Dict, List, Optional, Union
 
 import numpy as np
 import pandas as pd
@@ -146,7 +146,7 @@ def labels2onehot(
 
 def output2proba(
     preds: torch.Tensor, classes: OrderedSet, with_logits=False
-) -> Dict[ClfTarget, float]:
+) -> List[Dict[ClfTarget, float]]:
     if with_logits:
         if preds.shape[-1] >= 1:
             preds = torch.softmax(preds, dim=-1)
@@ -168,4 +168,4 @@ def output2proba(
         if preds_np.shape[0] == 1
         else [dict(zip(classes, pred)) for pred in preds_np]
     )
-    return dict(probas)
+    return [probas] if isinstance(probas, dict) else list(probas)

river_torch/utils/test_estimators.py → deep_river/utils/test_estimators.py

@@ -6,19 +6,19 @@
 import pytest
 import river
 
-from river_torch import utils
+from deep_river import utils
 
 
 def iter_estimators():
-    for submodule in importlib.import_module("river_torch").__all__:
+    for submodule in importlib.import_module("deep_river").__all__:
 
         def is_estimator(obj):
             return inspect.isclass(obj) and issubclass(
                 obj, river.base.Estimator
             )
 
         for _, obj in inspect.getmembers(
-            importlib.import_module(f"river_torch.{submodule}"), is_estimator
+            importlib.import_module(f"deep_river.{submodule}"), is_estimator
         ):
             yield obj