Note
Click here to download the full example code
Tensorflow/Keras#
Note
This example requires the tensorflow package to be installed.
Theoretically, tpcp is framework agnostic and can be used with any framework. However, due to the way some frameworks handle their objects, some special handling internally is required. Hence, this example does not only serve as example on how to use tensorflow with tpcp, but also as a test case for these special cases.
When using tpcp with any machine learning framework, you either want to use a pretrained model with a normal pipeline or a train your own model as part of an Optimizable Pipeline. Here we show the second case, as it is more complex, and you are likely able to figure out the first case yourself.
This means, we are planning to perform the following steps:
Create a pipeline that creates and trains a model.
Allow the modification of model hyperparameters.
Run a simple cross-validation to demonstrate the functionality.
This example reimplements the basic MNIST example from the [tensorflow documentation](https://www.tensorflow.org/tutorials/keras/classification).
Some Notes#
In this example we show how to implement a Pipeline that uses tensorflow. You could implement an Algorithm in a similar way. This would actually be easier, as no specific handling of the input data would be required. For a pipeline, we need to create a custom Dataset class, as this is the expected input for a pipeline.
The Dataset#
We are using the normal fashion MNIST dataset for this example It consists of 60.000 images of 28x28 pixels, each with a label. We will ignore the typical train-test split, as we want to do our own cross-validation.
In addition, we will simulate an additional “index level”. In this (and most typical deep learning datasets), each datapoint is one vector for which we can make one prediction. In tpcp, we usually deal with datasets, where you might have multiple pieces of information for each datapoint. For example, one datapoint could be a patient, for which we have an entire time series of measurements. We will simulate this here, by creating the index of our dataset as 1000 groups each containing 60 images.
Other than that, the dataset is pretty standard.
Besides the create_index method, we only need to implement the input_as_array and labels_as_array methods that
allow us to easily access the data once we selected a single group.
from functools import lru_cache
import numpy as np
import pandas as pd
import tensorflow as tf
from tpcp import Dataset
tf.keras.utils.set_random_seed(812)
tf.config.experimental.enable_op_determinism()
@lru_cache(maxsize=1)
def get_fashion_mnist_data():
# Note: We throw train and test sets together, as we don't care about the official split here.
# We will create our own split later.
(train_images, train_labels), (test_images, test_labels) = tf.keras.datasets.fashion_mnist.load_data()
return np.array(list(train_images) + list(test_images)), list(train_labels) + list(test_labels)
class FashionMNIST(Dataset):
def input_as_array(self) -> np.ndarray:
self.assert_is_single(None, "input_as_array")
group_id = int(self.group_label.group_id)
images, _ = get_fashion_mnist_data()
return images[group_id * 60 : (group_id + 1) * 60].reshape((60, 28, 28)) / 255
def labels_as_array(self) -> np.ndarray:
self.assert_is_single(None, "labels_as_array")
group_id = int(self.group_label.group_id)
_, labels = get_fashion_mnist_data()
return np.array(labels[group_id * 60 : (group_id + 1) * 60])
def create_index(self) -> pd.DataFrame:
# There are 60.000 images in total.
# We simulate 1000 groups of 60 images each.
return pd.DataFrame({"group_id": list(range(1000))})
We can see our Dataset works as expected:
dataset = FashionMNIST()
dataset[0].input_as_array().shape
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz
8192/29515 [=======>......................] - ETA: 0s
29515/29515 [==============================] - 0s 1us/step
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz
8192/26421880 [..............................] - ETA: 0s
131072/26421880 [..............................] - ETA: 10s
1040384/26421880 [>.............................] - ETA: 2s
5292032/26421880 [=====>........................] - ETA: 0s
8986624/26421880 [=========>....................] - ETA: 0s
13639680/26421880 [==============>...............] - ETA: 0s
16785408/26421880 [==================>...........] - ETA: 0s
21331968/26421880 [=======================>......] - ETA: 0s
26058752/26421880 [============================>.] - ETA: 0s
26421880/26421880 [==============================] - 0s 0us/step
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz
5148/5148 [==============================] - 0s 0us/step
Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz
8192/4422102 [..............................] - ETA: 0s
139264/4422102 [..............................] - ETA: 1s
1114112/4422102 [======>.......................] - ETA: 0s
4422102/4422102 [==============================] - 0s 0us/step
(60, 28, 28)
dataset[0].labels_as_array().shape
(60,)
The Pipeline#
We will create a pipeline that uses a simple neural network to classify the images.
In tpcp, all “things” that should be optimized need to be parameters.
This means our model itself needs to be a parameter of the pipeline.
However, as we don’t have the model yet, as its creation depends on other hyperparameters, we add it as an optional
parameter initialized with None.
Further, we prefix the parameter name with an underscore, to signify, that this is not a parameter that should be
modified manually by the user.
This is just convention, and it is up to you to decide how you want to name your parameters.
We further introduce a hyperparameter n_dense_layer_nodes to show how we can influence the model creation.
The optimize method#
To make our pipeline optimizable, it needs to inherit from OptimizablePipeline.
Further we need to mark at least one of the parameters as OptiPara using the type annotation.
We do this for our _model parameter.
Finally, we need to implement the self_optimize method.
This method will get the entire training dataset as input and should update the _model parameter with the trained
model.
Hence, we first extract the relevant data (remember, each datapoint is 60 images), by concatinating all images over
all groups in the dataset.
Then we create the Keras model based on the hyperparameters.
Finally, we train the model and update the _model parameter.
Here we chose to wrap the method with make_optimize_safe.
This decorator will perform some runtime checks to ensure that the method is implemented correctly.
The run method#
The run method expects that the _model parameter is already set (i.e. the pipeline was already optimized).
It gets a single datapoint as input (remember, a datapoint is a single group of 60 images).
We then extract the data from the datapoint and let the model make a prediction.
We store the prediction on our output attribute predictions_.
The trailing underscore is a convention to signify, that this is an “result” attribute.
from tpcp import OptimizablePipeline, OptiPara, make_optimize_safe, make_action_safe
from typing import Optional, Tuple
from typing_extensions import Self
import warnings
class KerasPipeline(OptimizablePipeline):
n_dense_layer_nodes: int
n_train_epochs: int
_model: OptiPara[Optional[tf.keras.Sequential]]
predictions_: np.ndarray
def __init__(self, n_dense_layer_nodes=128, n_train_epochs=5, _model: Optional[tf.keras.Sequential] = None):
self.n_dense_layer_nodes = n_dense_layer_nodes
self.n_train_epochs = n_train_epochs
self._model = _model
@property
def predicted_labels_(self):
return np.argmax(self.predictions_, axis=1)
@make_optimize_safe
def self_optimize(self, dataset, **_) -> Self:
data = np.vstack([d.input_as_array() for d in dataset])
labels = np.hstack([d.labels_as_array() for d in dataset])
print(data.shape)
if self._model is not None:
warnings.warn("Overwriting existing model!")
self._model = tf.keras.Sequential(
[
tf.keras.layers.Flatten(input_shape=(28, 28)),
tf.keras.layers.Dense(self.n_dense_layer_nodes, activation="relu"),
tf.keras.layers.Dense(10),
]
)
self._model.compile(
optimizer="adam",
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=["accuracy"],
)
self._model.fit(data, labels, epochs=self.n_train_epochs)
return self
@make_action_safe
def run(self, datapoint) -> Self:
if self._model is None:
raise RuntimeError("Model not trained yet!")
data = datapoint.input_as_array()
self.predictions_ = self._model.predict(data)
return self
Testing the pipeline#
We can now test our pipeline.
We will run the optimization using a couple of datapoints (to keep everything fast) and then use run to get the
predictions for a single unseen datapoint.
pipeline = KerasPipeline().self_optimize(FashionMNIST()[:10])
p1 = pipeline.run(FashionMNIST()[11])
print(p1.predicted_labels_)
print(FashionMNIST()[11].labels_as_array())
(600, 28, 28)
Epoch 1/5
1/19 [>.............................] - ETA: 12s - loss: 2.3593 - accuracy: 0.1250
19/19 [==============================] - 1s 2ms/step - loss: 1.5112 - accuracy: 0.4933
Epoch 2/5
1/19 [>.............................] - ETA: 0s - loss: 0.9220 - accuracy: 0.7188
19/19 [==============================] - 0s 2ms/step - loss: 0.8705 - accuracy: 0.7083
Epoch 3/5
1/19 [>.............................] - ETA: 0s - loss: 1.0179 - accuracy: 0.6875
19/19 [==============================] - 0s 2ms/step - loss: 0.6799 - accuracy: 0.7850
Epoch 4/5
1/19 [>.............................] - ETA: 0s - loss: 0.4521 - accuracy: 0.8750
19/19 [==============================] - 0s 2ms/step - loss: 0.5962 - accuracy: 0.8133
Epoch 5/5
1/19 [>.............................] - ETA: 0s - loss: 0.4230 - accuracy: 0.8438
19/19 [==============================] - 0s 2ms/step - loss: 0.5158 - accuracy: 0.8400
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
[8 8 0 9 6 0 7 3 7 9 3 8 6 3 7 8 1 4 0 7 9 8 5 5 2 1 3 3 1 9 7 5 9 9 7 8 2
7 2 7 2 6 7 1 1 7 5 4 8 3 5 9 0 7 3 0 0 9 1 9]
[8 8 0 9 2 0 7 3 7 9 3 8 4 3 7 8 1 4 0 7 9 8 5 5 2 1 3 4 6 7 7 5 9 9 7 8 2
7 4 7 0 3 5 1 1 5 5 2 8 3 5 9 0 7 3 0 0 7 1 9]
We can see that even with just 5 epochs, the model already performs quite well. To quantify we can calculate the accuracy for this datapoint:
from sklearn.metrics import accuracy_score
accuracy_score(p1.predicted_labels_, FashionMNIST()[11].labels_as_array())
0.8
Cross Validation#
If we want to run a cross validation, we need to formalize the scoring into a function. We will calculate two types of accuracy: First, the accuracy per group and second, the accuracy over all images across all groups. For more information about how this works, check the custom_scorer example.
from typing import Sequence, Dict
from tpcp.validate import Aggregator
class SingleValueAccuracy(Aggregator[np.ndarray]):
RETURN_RAW_SCORES = False
@classmethod
def aggregate(cls, /, values: Sequence[Tuple[np.ndarray, np.ndarray]], **_) -> Dict[str, float]:
return {"accuracy": accuracy_score(np.hstack([v[0] for v in values]), np.hstack([v[1] for v in values]))}
def scoring(pipeline, datapoint):
result: np.ndarray = pipeline.safe_run(datapoint).predicted_labels_
reference = datapoint.labels_as_array()
return {
"accuracy": accuracy_score(result, reference),
"per_sample": SingleValueAccuracy((result, reference)),
}
Now we can run a cross validation. We will only run it on a subset of the data, to keep the runtime manageable.
Note
You might see warnings about retracing of the model. This is because we clone the pipeline before each call to the run method. This is a good idea to ensure that all pipelines are independent of each other, however, might result in some performance overhead.
from tpcp.validate import cross_validate
from tpcp.optimize import Optimize
pipeline = KerasPipeline(n_train_epochs=10)
cv_results = cross_validate(Optimize(pipeline), FashionMNIST()[:100], scoring=scoring, cv=3)
CV Folds: 0%| | 0/3 [00:00<?, ?it/s](3960, 28, 28)
Epoch 1/10
1/124 [..............................] - ETA: 1:12 - loss: 2.5074 - accuracy: 0.0938
25/124 [=====>........................] - ETA: 0s - loss: 1.4785 - accuracy: 0.4600
50/124 [===========>..................] - ETA: 0s - loss: 1.1601 - accuracy: 0.5863
75/124 [=================>............] - ETA: 0s - loss: 1.0235 - accuracy: 0.6354
100/124 [=======================>......] - ETA: 0s - loss: 0.9496 - accuracy: 0.6678
124/124 [==============================] - 1s 2ms/step - loss: 0.8968 - accuracy: 0.6919
Epoch 2/10
1/124 [..............................] - ETA: 0s - loss: 1.0481 - accuracy: 0.7500
26/124 [=====>........................] - ETA: 0s - loss: 0.5908 - accuracy: 0.8089
51/124 [===========>..................] - ETA: 0s - loss: 0.5920 - accuracy: 0.7996
76/124 [=================>............] - ETA: 0s - loss: 0.5642 - accuracy: 0.8076
101/124 [=======================>......] - ETA: 0s - loss: 0.5612 - accuracy: 0.8103
124/124 [==============================] - 0s 2ms/step - loss: 0.5606 - accuracy: 0.8104
Epoch 3/10
1/124 [..............................] - ETA: 0s - loss: 0.7180 - accuracy: 0.6250
26/124 [=====>........................] - ETA: 0s - loss: 0.5136 - accuracy: 0.8161
51/124 [===========>..................] - ETA: 0s - loss: 0.5116 - accuracy: 0.8241
76/124 [=================>............] - ETA: 0s - loss: 0.4994 - accuracy: 0.8240
101/124 [=======================>......] - ETA: 0s - loss: 0.4985 - accuracy: 0.8261
124/124 [==============================] - 0s 2ms/step - loss: 0.5009 - accuracy: 0.8232
Epoch 4/10
1/124 [..............................] - ETA: 0s - loss: 0.4630 - accuracy: 0.8125
26/124 [=====>........................] - ETA: 0s - loss: 0.4043 - accuracy: 0.8618
51/124 [===========>..................] - ETA: 0s - loss: 0.4236 - accuracy: 0.8529
76/124 [=================>............] - ETA: 0s - loss: 0.4401 - accuracy: 0.8470
101/124 [=======================>......] - ETA: 0s - loss: 0.4450 - accuracy: 0.8450
124/124 [==============================] - 0s 2ms/step - loss: 0.4448 - accuracy: 0.8437
Epoch 5/10
1/124 [..............................] - ETA: 0s - loss: 0.6924 - accuracy: 0.6250
26/124 [=====>........................] - ETA: 0s - loss: 0.4356 - accuracy: 0.8281
51/124 [===========>..................] - ETA: 0s - loss: 0.4340 - accuracy: 0.8407
76/124 [=================>............] - ETA: 0s - loss: 0.4171 - accuracy: 0.8503
101/124 [=======================>......] - ETA: 0s - loss: 0.4101 - accuracy: 0.8555
124/124 [==============================] - 0s 2ms/step - loss: 0.4119 - accuracy: 0.8540
Epoch 6/10
1/124 [..............................] - ETA: 0s - loss: 0.5976 - accuracy: 0.7812
26/124 [=====>........................] - ETA: 0s - loss: 0.4092 - accuracy: 0.8558
51/124 [===========>..................] - ETA: 0s - loss: 0.3851 - accuracy: 0.8664
76/124 [=================>............] - ETA: 0s - loss: 0.3845 - accuracy: 0.8660
101/124 [=======================>......] - ETA: 0s - loss: 0.3790 - accuracy: 0.8691
124/124 [==============================] - 0s 2ms/step - loss: 0.3792 - accuracy: 0.8687
Epoch 7/10
1/124 [..............................] - ETA: 0s - loss: 0.4009 - accuracy: 0.8438
26/124 [=====>........................] - ETA: 0s - loss: 0.3543 - accuracy: 0.8714
51/124 [===========>..................] - ETA: 0s - loss: 0.3607 - accuracy: 0.8713
76/124 [=================>............] - ETA: 0s - loss: 0.3589 - accuracy: 0.8758
101/124 [=======================>......] - ETA: 0s - loss: 0.3569 - accuracy: 0.8790
124/124 [==============================] - 0s 2ms/step - loss: 0.3555 - accuracy: 0.8768
Epoch 8/10
1/124 [..............................] - ETA: 0s - loss: 0.2810 - accuracy: 0.9062
26/124 [=====>........................] - ETA: 0s - loss: 0.2951 - accuracy: 0.9147
51/124 [===========>..................] - ETA: 0s - loss: 0.3004 - accuracy: 0.9038
76/124 [=================>............] - ETA: 0s - loss: 0.3195 - accuracy: 0.8935
101/124 [=======================>......] - ETA: 0s - loss: 0.3299 - accuracy: 0.8899
124/124 [==============================] - 0s 2ms/step - loss: 0.3464 - accuracy: 0.8826
Epoch 9/10
1/124 [..............................] - ETA: 0s - loss: 0.2351 - accuracy: 0.9375
27/124 [=====>........................] - ETA: 0s - loss: 0.3445 - accuracy: 0.8843
52/124 [===========>..................] - ETA: 0s - loss: 0.3397 - accuracy: 0.8792
77/124 [=================>............] - ETA: 0s - loss: 0.3213 - accuracy: 0.8884
102/124 [=======================>......] - ETA: 0s - loss: 0.3224 - accuracy: 0.8903
124/124 [==============================] - 0s 2ms/step - loss: 0.3196 - accuracy: 0.8886
Epoch 10/10
1/124 [..............................] - ETA: 0s - loss: 0.2527 - accuracy: 0.9688
26/124 [=====>........................] - ETA: 0s - loss: 0.2442 - accuracy: 0.9111
51/124 [===========>..................] - ETA: 0s - loss: 0.2760 - accuracy: 0.9044
76/124 [=================>............] - ETA: 0s - loss: 0.2979 - accuracy: 0.8960
101/124 [=======================>......] - ETA: 0s - loss: 0.2960 - accuracy: 0.8982
124/124 [==============================] - 0s 2ms/step - loss: 0.2921 - accuracy: 0.8997
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
CV Folds: 33%|###3 | 1/3 [00:11<00:22, 11.48s/it](4020, 28, 28)
Epoch 1/10
1/126 [..............................] - ETA: 1:12 - loss: 2.3656 - accuracy: 0.1250
26/126 [=====>........................] - ETA: 0s - loss: 1.4324 - accuracy: 0.5240
51/126 [===========>..................] - ETA: 0s - loss: 1.1858 - accuracy: 0.6048
76/126 [=================>............] - ETA: 0s - loss: 1.0536 - accuracy: 0.6460
100/126 [======================>.......] - ETA: 0s - loss: 0.9618 - accuracy: 0.6753
125/126 [============================>.] - ETA: 0s - loss: 0.8932 - accuracy: 0.6990
126/126 [==============================] - 1s 2ms/step - loss: 0.8913 - accuracy: 0.7000
Epoch 2/10
1/126 [..............................] - ETA: 0s - loss: 0.4984 - accuracy: 0.8125
26/126 [=====>........................] - ETA: 0s - loss: 0.5686 - accuracy: 0.8053
51/126 [===========>..................] - ETA: 0s - loss: 0.5829 - accuracy: 0.7923
76/126 [=================>............] - ETA: 0s - loss: 0.5839 - accuracy: 0.7940
101/126 [=======================>......] - ETA: 0s - loss: 0.5740 - accuracy: 0.7942
126/126 [==============================] - ETA: 0s - loss: 0.5662 - accuracy: 0.8002
126/126 [==============================] - 0s 2ms/step - loss: 0.5662 - accuracy: 0.8002
Epoch 3/10
1/126 [..............................] - ETA: 0s - loss: 0.5249 - accuracy: 0.8125
26/126 [=====>........................] - ETA: 0s - loss: 0.5084 - accuracy: 0.8317
51/126 [===========>..................] - ETA: 0s - loss: 0.4669 - accuracy: 0.8364
76/126 [=================>............] - ETA: 0s - loss: 0.4701 - accuracy: 0.8363
101/126 [=======================>......] - ETA: 0s - loss: 0.4785 - accuracy: 0.8348
126/126 [==============================] - ETA: 0s - loss: 0.4893 - accuracy: 0.8279
126/126 [==============================] - 0s 2ms/step - loss: 0.4893 - accuracy: 0.8279
Epoch 4/10
1/126 [..............................] - ETA: 0s - loss: 0.2082 - accuracy: 0.9688
26/126 [=====>........................] - ETA: 0s - loss: 0.4251 - accuracy: 0.8594
51/126 [===========>..................] - ETA: 0s - loss: 0.4422 - accuracy: 0.8456
76/126 [=================>............] - ETA: 0s - loss: 0.4264 - accuracy: 0.8536
101/126 [=======================>......] - ETA: 0s - loss: 0.4435 - accuracy: 0.8478
126/126 [==============================] - ETA: 0s - loss: 0.4431 - accuracy: 0.8453
126/126 [==============================] - 0s 2ms/step - loss: 0.4431 - accuracy: 0.8453
Epoch 5/10
1/126 [..............................] - ETA: 0s - loss: 0.1979 - accuracy: 0.9375
26/126 [=====>........................] - ETA: 0s - loss: 0.4029 - accuracy: 0.8558
51/126 [===========>..................] - ETA: 0s - loss: 0.4061 - accuracy: 0.8548
76/126 [=================>............] - ETA: 0s - loss: 0.4097 - accuracy: 0.8553
101/126 [=======================>......] - ETA: 0s - loss: 0.4030 - accuracy: 0.8567
125/126 [============================>.] - ETA: 0s - loss: 0.4040 - accuracy: 0.8583
126/126 [==============================] - 0s 2ms/step - loss: 0.4046 - accuracy: 0.8580
Epoch 6/10
1/126 [..............................] - ETA: 0s - loss: 0.4598 - accuracy: 0.8125
25/126 [====>.........................] - ETA: 0s - loss: 0.3639 - accuracy: 0.8625
50/126 [==========>...................] - ETA: 0s - loss: 0.3693 - accuracy: 0.8637
75/126 [================>.............] - ETA: 0s - loss: 0.3757 - accuracy: 0.8642
100/126 [======================>.......] - ETA: 0s - loss: 0.3760 - accuracy: 0.8659
125/126 [============================>.] - ETA: 0s - loss: 0.3771 - accuracy: 0.8670
126/126 [==============================] - 0s 2ms/step - loss: 0.3772 - accuracy: 0.8667
Epoch 7/10
1/126 [..............................] - ETA: 0s - loss: 0.2244 - accuracy: 0.9062
25/126 [====>.........................] - ETA: 0s - loss: 0.3532 - accuracy: 0.8813
50/126 [==========>...................] - ETA: 0s - loss: 0.3309 - accuracy: 0.8875
75/126 [================>.............] - ETA: 0s - loss: 0.3474 - accuracy: 0.8813
100/126 [======================>.......] - ETA: 0s - loss: 0.3482 - accuracy: 0.8794
125/126 [============================>.] - ETA: 0s - loss: 0.3489 - accuracy: 0.8780
126/126 [==============================] - 0s 2ms/step - loss: 0.3479 - accuracy: 0.8786
Epoch 8/10
1/126 [..............................] - ETA: 0s - loss: 0.2092 - accuracy: 0.9375
25/126 [====>.........................] - ETA: 0s - loss: 0.2997 - accuracy: 0.9000
50/126 [==========>...................] - ETA: 0s - loss: 0.3115 - accuracy: 0.8931
75/126 [================>.............] - ETA: 0s - loss: 0.3209 - accuracy: 0.8888
100/126 [======================>.......] - ETA: 0s - loss: 0.3294 - accuracy: 0.8869
125/126 [============================>.] - ETA: 0s - loss: 0.3292 - accuracy: 0.8860
126/126 [==============================] - 0s 2ms/step - loss: 0.3285 - accuracy: 0.8863
Epoch 9/10
1/126 [..............................] - ETA: 0s - loss: 0.1886 - accuracy: 0.9062
26/126 [=====>........................] - ETA: 0s - loss: 0.2963 - accuracy: 0.9014
50/126 [==========>...................] - ETA: 0s - loss: 0.3077 - accuracy: 0.8881
75/126 [================>.............] - ETA: 0s - loss: 0.3219 - accuracy: 0.8771
100/126 [======================>.......] - ETA: 0s - loss: 0.3062 - accuracy: 0.8866
125/126 [============================>.] - ETA: 0s - loss: 0.3130 - accuracy: 0.8898
126/126 [==============================] - 0s 2ms/step - loss: 0.3120 - accuracy: 0.8900
Epoch 10/10
1/126 [..............................] - ETA: 0s - loss: 0.2327 - accuracy: 0.9062
26/126 [=====>........................] - ETA: 0s - loss: 0.2941 - accuracy: 0.8822
51/126 [===========>..................] - ETA: 0s - loss: 0.2947 - accuracy: 0.8922
76/126 [=================>............] - ETA: 0s - loss: 0.3047 - accuracy: 0.8882
101/126 [=======================>......] - ETA: 0s - loss: 0.3064 - accuracy: 0.8877
126/126 [==============================] - ETA: 0s - loss: 0.3065 - accuracy: 0.8871
126/126 [==============================] - 0s 2ms/step - loss: 0.3065 - accuracy: 0.8871
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
CV Folds: 67%|######6 | 2/3 [00:22<00:11, 11.40s/it](4020, 28, 28)
Epoch 1/10
1/126 [..............................] - ETA: 1:13 - loss: 2.3129 - accuracy: 0.0938
25/126 [====>.........................] - ETA: 0s - loss: 1.4511 - accuracy: 0.5113
50/126 [==========>...................] - ETA: 0s - loss: 1.1892 - accuracy: 0.5994
75/126 [================>.............] - ETA: 0s - loss: 1.0642 - accuracy: 0.6392
100/126 [======================>.......] - ETA: 0s - loss: 0.9617 - accuracy: 0.6753
125/126 [============================>.] - ETA: 0s - loss: 0.9002 - accuracy: 0.6967
126/126 [==============================] - 1s 2ms/step - loss: 0.8965 - accuracy: 0.6983
Epoch 2/10
1/126 [..............................] - ETA: 0s - loss: 0.4769 - accuracy: 0.8125
26/126 [=====>........................] - ETA: 0s - loss: 0.6113 - accuracy: 0.7957
51/126 [===========>..................] - ETA: 0s - loss: 0.5969 - accuracy: 0.7953
76/126 [=================>............] - ETA: 0s - loss: 0.5938 - accuracy: 0.7956
101/126 [=======================>......] - ETA: 0s - loss: 0.5780 - accuracy: 0.8026
126/126 [==============================] - ETA: 0s - loss: 0.5750 - accuracy: 0.8040
126/126 [==============================] - 0s 2ms/step - loss: 0.5750 - accuracy: 0.8040
Epoch 3/10
1/126 [..............................] - ETA: 0s - loss: 0.6893 - accuracy: 0.7812
25/126 [====>.........................] - ETA: 0s - loss: 0.5681 - accuracy: 0.8075
50/126 [==========>...................] - ETA: 0s - loss: 0.5351 - accuracy: 0.8125
75/126 [================>.............] - ETA: 0s - loss: 0.5261 - accuracy: 0.8179
100/126 [======================>.......] - ETA: 0s - loss: 0.5193 - accuracy: 0.8222
125/126 [============================>.] - ETA: 0s - loss: 0.5144 - accuracy: 0.8227
126/126 [==============================] - 0s 2ms/step - loss: 0.5140 - accuracy: 0.8224
Epoch 4/10
1/126 [..............................] - ETA: 0s - loss: 0.4519 - accuracy: 0.9062
26/126 [=====>........................] - ETA: 0s - loss: 0.4196 - accuracy: 0.8582
51/126 [===========>..................] - ETA: 0s - loss: 0.4550 - accuracy: 0.8395
76/126 [=================>............] - ETA: 0s - loss: 0.4425 - accuracy: 0.8446
101/126 [=======================>......] - ETA: 0s - loss: 0.4570 - accuracy: 0.8416
126/126 [==============================] - ETA: 0s - loss: 0.4667 - accuracy: 0.8358
126/126 [==============================] - 0s 2ms/step - loss: 0.4667 - accuracy: 0.8358
Epoch 5/10
1/126 [..............................] - ETA: 0s - loss: 0.4192 - accuracy: 0.8125
25/126 [====>.........................] - ETA: 0s - loss: 0.4182 - accuracy: 0.8625
50/126 [==========>...................] - ETA: 0s - loss: 0.4171 - accuracy: 0.8656
75/126 [================>.............] - ETA: 0s - loss: 0.4220 - accuracy: 0.8608
100/126 [======================>.......] - ETA: 0s - loss: 0.4263 - accuracy: 0.8584
125/126 [============================>.] - ETA: 0s - loss: 0.4181 - accuracy: 0.8597
126/126 [==============================] - 0s 2ms/step - loss: 0.4173 - accuracy: 0.8600
Epoch 6/10
1/126 [..............................] - ETA: 0s - loss: 0.8799 - accuracy: 0.7500
26/126 [=====>........................] - ETA: 0s - loss: 0.3717 - accuracy: 0.8654
51/126 [===========>..................] - ETA: 0s - loss: 0.3801 - accuracy: 0.8701
76/126 [=================>............] - ETA: 0s - loss: 0.3856 - accuracy: 0.8688
101/126 [=======================>......] - ETA: 0s - loss: 0.3907 - accuracy: 0.8632
126/126 [==============================] - ETA: 0s - loss: 0.3949 - accuracy: 0.8624
126/126 [==============================] - 0s 2ms/step - loss: 0.3949 - accuracy: 0.8624
Epoch 7/10
1/126 [..............................] - ETA: 0s - loss: 0.2022 - accuracy: 0.9375
26/126 [=====>........................] - ETA: 0s - loss: 0.3734 - accuracy: 0.8786
51/126 [===========>..................] - ETA: 0s - loss: 0.3456 - accuracy: 0.8817
76/126 [=================>............] - ETA: 0s - loss: 0.3615 - accuracy: 0.8762
101/126 [=======================>......] - ETA: 0s - loss: 0.3618 - accuracy: 0.8769
126/126 [==============================] - ETA: 0s - loss: 0.3635 - accuracy: 0.8774
126/126 [==============================] - 0s 2ms/step - loss: 0.3635 - accuracy: 0.8774
Epoch 8/10
1/126 [..............................] - ETA: 0s - loss: 0.3186 - accuracy: 0.8750
26/126 [=====>........................] - ETA: 0s - loss: 0.3445 - accuracy: 0.8810
50/126 [==========>...................] - ETA: 0s - loss: 0.3372 - accuracy: 0.8844
74/126 [================>.............] - ETA: 0s - loss: 0.3362 - accuracy: 0.8847
98/126 [======================>.......] - ETA: 0s - loss: 0.3342 - accuracy: 0.8865
123/126 [============================>.] - ETA: 0s - loss: 0.3448 - accuracy: 0.8824
126/126 [==============================] - 0s 2ms/step - loss: 0.3457 - accuracy: 0.8826
Epoch 9/10
1/126 [..............................] - ETA: 0s - loss: 0.2176 - accuracy: 0.9375
26/126 [=====>........................] - ETA: 0s - loss: 0.3434 - accuracy: 0.8834
50/126 [==========>...................] - ETA: 0s - loss: 0.3422 - accuracy: 0.8750
75/126 [================>.............] - ETA: 0s - loss: 0.3323 - accuracy: 0.8804
100/126 [======================>.......] - ETA: 0s - loss: 0.3298 - accuracy: 0.8853
125/126 [============================>.] - ETA: 0s - loss: 0.3278 - accuracy: 0.8880
126/126 [==============================] - 0s 2ms/step - loss: 0.3285 - accuracy: 0.8878
Epoch 10/10
1/126 [..............................] - ETA: 0s - loss: 0.3079 - accuracy: 0.8125
26/126 [=====>........................] - ETA: 0s - loss: 0.2803 - accuracy: 0.9075
51/126 [===========>..................] - ETA: 0s - loss: 0.2911 - accuracy: 0.9044
76/126 [=================>............] - ETA: 0s - loss: 0.2854 - accuracy: 0.9062
100/126 [======================>.......] - ETA: 0s - loss: 0.2977 - accuracy: 0.9016
125/126 [============================>.] - ETA: 0s - loss: 0.3012 - accuracy: 0.8992
126/126 [==============================] - 0s 2ms/step - loss: 0.3008 - accuracy: 0.8993
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 2ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
1/2 [==============>...............] - ETA: 0s
2/2 [==============================] - 0s 1ms/step
CV Folds: 100%|##########| 3/3 [00:32<00:00, 10.41s/it]
CV Folds: 100%|##########| 3/3 [00:32<00:00, 10.69s/it]
We can now look at the results per group:
cv_results["test_single_accuracy"]
[[0.85, 0.8666666666666667, 0.9333333333333333, 0.8, 0.85, 0.85, 0.85, 0.8833333333333333, 0.85, 0.85, 0.8166666666666667, 0.8666666666666667, 0.9333333333333333, 0.8, 0.8833333333333333, 0.8, 0.8, 0.8833333333333333, 0.7833333333333333, 0.8166666666666667, 0.8, 0.85, 0.8, 0.95, 0.8, 0.8833333333333333, 0.8833333333333333, 0.85, 0.85, 0.7833333333333333, 0.75, 0.8666666666666667, 0.8333333333333334, 0.9333333333333333], [0.7833333333333333, 0.8166666666666667, 0.7666666666666667, 0.7833333333333333, 0.85, 0.8, 0.8, 0.8166666666666667, 0.8, 0.8333333333333334, 0.8333333333333334, 0.8, 0.8166666666666667, 0.85, 0.75, 0.7, 0.8333333333333334, 0.8166666666666667, 0.8666666666666667, 0.7666666666666667, 0.7833333333333333, 0.7833333333333333, 0.7666666666666667, 0.8833333333333333, 0.8333333333333334, 0.7833333333333333, 0.8, 0.85, 0.85, 0.8166666666666667, 0.7666666666666667, 0.8, 0.8833333333333333], [0.7666666666666667, 0.9166666666666666, 0.8, 0.8, 0.7833333333333333, 0.85, 0.8833333333333333, 0.85, 0.9, 0.8833333333333333, 0.9, 0.85, 0.8833333333333333, 0.8166666666666667, 0.8333333333333334, 0.8333333333333334, 0.85, 0.7833333333333333, 0.7166666666666667, 0.7333333333333333, 0.75, 0.7833333333333333, 0.85, 0.7666666666666667, 0.7833333333333333, 0.9, 0.8666666666666667, 0.8333333333333334, 0.8333333333333334, 0.8166666666666667, 0.85, 0.85, 0.9]]
And the overall accuracy as the average over all samples of all groups within a fold:
cv_results["test_per_sample__accuracy"]
array([0.84705882, 0.80858586, 0.83080808])
Total running time of the script: ( 0 minutes 36.573 seconds)
Estimated memory usage: 253 MB