diff --git a/docs/tutorials/convolutional_neural_network_for_image_classification.ipynb b/docs/tutorials/convolutional_neural_network_for_image_classification.ipynb index ee15eaa03..2c06928f5 100644 --- a/docs/tutorials/convolutional_neural_network_for_image_classification.ipynb +++ b/docs/tutorials/convolutional_neural_network_for_image_classification.ipynb @@ -2,14 +2,18 @@ "cells": [ { "cell_type": "markdown", - "source": [], + "id": "6d2c03ea985891aa", "metadata": { "collapsed": false }, - "id": "6d2c03ea985891aa" + "source": [] }, { "cell_type": "markdown", + "id": "d65ab4a80f6bc842", + "metadata": { + "collapsed": false + }, "source": [ "# Image Classification with Convolutional Neural Networks\n", "\n", @@ -21,119 +25,125 @@ " All operations on an NeuralNetworkClassifier return a new NeuralNetworkClassifier. The original NeuralNetworkClassifier will not be changed.\n", "

\n", "" - ], - "metadata": { - "collapsed": false - }, - "id": "d65ab4a80f6bc842" + ] }, { "cell_type": "markdown", - "source": "## Load data into an `ImageDataset`", + "id": "74dacfa56deeaed3", "metadata": { "collapsed": false }, - "id": "74dacfa56deeaed3" + "source": [ + "## Load data into an `ImageDataset`" + ] }, { "cell_type": "markdown", - "source": [ - "1. Load images via files in an `ImageList`. The data is available under `docs/tutorials/data/shapes`. If the `return_filenames` parameter is set to `True`, a list of all filepaths will be returned as well in the same order as the images in the returned `ImageList`." - ], + "id": "90dfbc18037f0201", "metadata": { "collapsed": false }, - "id": "90dfbc18037f0201" + "source": [ + "1. Load images via files in an `ImageList`. The data is available under `docs/tutorials/data/shapes`. If the `return_filenames` parameter is set to `True`, a list of all filepaths will be returned as well in the same order as the images in the returned `ImageList`." + ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 1, "id": "initial_id", "metadata": { "collapsed": true }, + "outputs": [], "source": [ "from safeds.data.image.containers import ImageList\n", "\n", "images, filepaths = ImageList.from_files(\"data/shapes\", return_filenames=True)" - ], - "outputs": [] + ] }, { "cell_type": "markdown", - "source": [ - "2. Create a `Column` with the labels of the images:" - ], + "id": "76bc8612f449edf", "metadata": { "collapsed": false }, - "id": "76bc8612f449edf" + "source": [ + "2. Create a `Column` with the labels of the images:" + ] }, { "cell_type": "code", + "execution_count": null, + "id": "66dcf95a3fa51f23", + "metadata": { + "collapsed": false + }, + "outputs": [], "source": [ "import re\n", + "\n", "from safeds.data.tabular.containers import Column\n", "\n", "labels = Column(\n", " \"label\", \n", - " [re.search(r\"(.*)[\\\\/](.*)[\\\\/](.*)\\.\", filepath).group(2) for filepath in filepaths]\n", + " [re.search(r\"(.*)[\\\\/](.*)[\\\\/](.*)\\.\", filepath).group(2) for filepath in filepaths],\n", ")" - ], - "metadata": { - "collapsed": false - }, - "id": "66dcf95a3fa51f23", - "execution_count": null, - "outputs": [] + ] }, { "cell_type": "markdown", - "source": [ - "3. Create an `ImageDataset` from the `ImageList` and the `Column` of labels. If the `shuffle` parameter is set to `True`, the `ImageDataset` will be shuffled after each epoch while training a neural network." - ], + "id": "596b0c9ec9627ad0", "metadata": { "collapsed": false }, - "id": "596b0c9ec9627ad0" + "source": [ + "3. Create an `ImageDataset` from the `ImageList` and the `Column` of labels. If the `shuffle` parameter is set to `True`, the `ImageDataset` will be shuffled after each epoch while training a neural network." + ] }, { "cell_type": "code", + "execution_count": null, + "id": "32056ddf5396e070", + "metadata": { + "collapsed": false + }, + "outputs": [], "source": [ "from safeds.data.labeled.containers import ImageDataset\n", "\n", "dataset = ImageDataset[Column](images, labels, shuffle=True)" - ], - "metadata": { - "collapsed": false - }, - "id": "32056ddf5396e070", - "execution_count": null, - "outputs": [] + ] }, { "cell_type": "markdown", - "source": "## Create the neural network with a `NeuralNetworkClassifier`", + "id": "358bd4cc05c8daf3", "metadata": { "collapsed": false }, - "id": "358bd4cc05c8daf3" + "source": [ + "## Create the neural network with a `NeuralNetworkClassifier`" + ] }, { "cell_type": "markdown", - "source": [ - "1. Create a list of `Layer` instances for your neural network:" - ], + "id": "fe40c93a1cfd3a7b", "metadata": { "collapsed": false }, - "id": "fe40c93a1cfd3a7b" + "source": [ + "1. Create a list of `Layer` instances for your neural network:" + ] }, { "cell_type": "code", + "execution_count": null, + "id": "806a8091249d533a", + "metadata": { + "collapsed": false + }, + "outputs": [], "source": [ - "from safeds.ml.nn.layers import (Convolutional2DLayer, FlattenLayer,\n", - " ForwardLayer, MaxPooling2DLayer)\n", + "from safeds.ml.nn.layers import Convolutional2DLayer, FlattenLayer, ForwardLayer, MaxPooling2DLayer\n", "\n", "layers = [\n", " Convolutional2DLayer(output_channel=16, kernel_size=3, padding=1),\n", @@ -144,24 +154,26 @@ " ForwardLayer(neuron_count=128),\n", " ForwardLayer(neuron_count=3),\n", "]" - ], - "metadata": { - "collapsed": false - }, - "id": "806a8091249d533a", - "execution_count": null, - "outputs": [] + ] }, { "cell_type": "markdown", - "source": "2. Create a `NeuralNetworkClassifier` from an `InputConversion`, the list of `Layer` instances:", + "id": "fe4f6a4d14404a85", "metadata": { "collapsed": false }, - "id": "fe4f6a4d14404a85" + "source": [ + "2. Create a `NeuralNetworkClassifier` from an `InputConversion`, the list of `Layer` instances:" + ] }, { "cell_type": "code", + "execution_count": null, + "id": "af68cc0d32655d32", + "metadata": { + "collapsed": false + }, + "outputs": [], "source": [ "from safeds.ml.nn import NeuralNetworkClassifier\n", "from safeds.ml.nn.converters import InputConversionImageToColumn\n", @@ -170,131 +182,158 @@ " InputConversionImageToColumn(dataset.input_size), \n", " layers,\n", ")" - ], - "metadata": { - "collapsed": false - }, - "id": "af68cc0d32655d32", - "execution_count": null, - "outputs": [] + ] }, { "cell_type": "markdown", - "source": "## Fit and predict the `NeuralNetworkClassifier`", + "id": "4f9387686ba50c37", "metadata": { "collapsed": false }, - "id": "4f9387686ba50c37" + "source": [ + "## Fit and predict the `NeuralNetworkClassifier`" + ] }, { "cell_type": "markdown", - "source": [ - "1. Fit the `NeuralNetworkClassifier`:" - ], + "id": "3d8efa74951725cb", "metadata": { "collapsed": false }, - "id": "3d8efa74951725cb" + "source": [ + "1. Fit the `NeuralNetworkClassifier`:" + ] }, { "cell_type": "code", - "source": [ - "cnn_fitted = cnn.fit(dataset, epoch_size=32, batch_size=16)" - ], + "execution_count": null, + "id": "381627a94d500675", "metadata": { "collapsed": false }, - "id": "381627a94d500675", - "execution_count": null, - "outputs": [] + "outputs": [], + "source": [ + "cnn_fitted = cnn.fit(dataset, epoch_size=32, batch_size=16)" + ] }, { "cell_type": "markdown", - "source": [ - "2. Predict values from the `NeuralNetworkClassifier`:" - ], + "id": "35bb7d0ebfabf597", "metadata": { "collapsed": false }, - "id": "35bb7d0ebfabf597" + "source": [ + "2. Predict values from the `NeuralNetworkClassifier`:" + ] }, { "cell_type": "code", - "source": [ - "prediction = cnn_fitted.predict(dataset.get_input())" - ], + "execution_count": null, + "id": "62f63dd68362c8b7", "metadata": { "collapsed": false }, - "id": "62f63dd68362c8b7", - "execution_count": null, - "outputs": [] + "outputs": [], + "source": [ + "prediction = cnn_fitted.predict(dataset.get_input())" + ] }, { "cell_type": "markdown", - "source": [ - "3. Shuffle the prediction to get a random order:" - ], + "id": "a8ecd71982a0cc97", "metadata": { "collapsed": false }, - "id": "a8ecd71982a0cc97" + "source": [ + "3. Shuffle the prediction to get a random order:" + ] }, { "cell_type": "code", - "source": [ - "shuffled_prediction = prediction.shuffle()" - ], + "execution_count": null, + "id": "779277d73e30554d", "metadata": { "collapsed": false }, - "id": "779277d73e30554d", - "execution_count": null, - "outputs": [] + "outputs": [], + "source": [ + "shuffled_prediction = prediction.shuffle()" + ] }, { "cell_type": "markdown", - "source": [ - "4. Display a subset of the input data:" - ], + "id": "2c1ae7438df15cae", "metadata": { "collapsed": false }, - "id": "2c1ae7438df15cae" + "source": [ + "4. Display a subset of the input data:" + ] }, { "cell_type": "code", - "source": [ - "shuffled_prediction.get_input().remove_image_by_index(list(range(9, len(prediction))))" - ], + "execution_count": null, + "id": "a5ddbbfba41aa7f", "metadata": { "collapsed": false }, - "id": "a5ddbbfba41aa7f", - "execution_count": null, - "outputs": [] + "outputs": [ + { + "data": { + "image/png": "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", + "text/plain": [ + "" + ] + }, + "execution_count": 10, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "shuffled_prediction.get_input().remove_image_by_index(list(range(9, len(prediction))))" + ] }, { "cell_type": "markdown", - "source": [ - "5. Display the corresponding predicted labels:" - ], + "id": "131db684a431d4ec", "metadata": { "collapsed": false }, - "id": "131db684a431d4ec" + "source": [ + "5. Display the corresponding predicted labels:" + ] }, { "cell_type": "code", - "source": [ - "shuffled_prediction.get_output().to_list()[0:9]" - ], + "execution_count": null, + "id": "7081595d7100fb42", "metadata": { "collapsed": false }, - "id": "7081595d7100fb42", - "execution_count": null, - "outputs": [] + "outputs": [ + { + "data": { + "text/plain": [ + "['triangles',\n", + " 'triangles',\n", + " 'triangles',\n", + " 'triangles',\n", + " 'circles',\n", + " 'squares',\n", + " 'triangles',\n", + " 'squares',\n", + " 'squares']" + ] + }, + "execution_count": 11, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "shuffled_prediction.get_output().to_list()[0:9]" + ] } ], "metadata": { @@ -306,14 +345,14 @@ "language_info": { "codemirror_mode": { "name": "ipython", - "version": 2 + "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", - "pygments_lexer": "ipython2", - "version": "2.7.6" + "pygments_lexer": "ipython3", + "version": "3.12.4" } }, "nbformat": 4, diff --git a/docs/tutorials/regression.ipynb b/docs/tutorials/regression.ipynb index 6ec3792a3..c966031f8 100644 --- a/docs/tutorials/regression.ipynb +++ b/docs/tutorials/regression.ipynb @@ -33,11 +33,48 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 1, "metadata": { "collapsed": false }, - "outputs": [], + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "shape: (15, 23)
idyearmonthdayzipcodelatitudelongitudesqft_lotsqft_livingsqft_abovesqft_basementfloorsbedroomsbathroomswaterfrontviewconditiongradeyear_builtyear_renovatedsqft_lot_15nnsqft_living_15nnprice
i64i64i64i64i64f64f64i64i64i64i64f64i64f64i64i64i64i64i64i64i64i64i64
02014529800147.3406-122.26993972200220002.042.501381987091762310285000
12014529800347.3537-122.30310834209013607301.032.501481987085951750285000
22014529800647.5443-122.17781192160108010801.042.2501381966090001850440000
32014529800647.5746-122.13588001450145001.041.001471954089421260435000
42014529800647.5725-122.13310000192010708501.041.5014719540108361450430000
……………………………………………………………
102014529802347.3256-122.378331513240324002.032.50331019950249674050604000
112014529802447.5643-121.897162151580158001.032.25014719780162151450335000
122014529802747.4635-121.991351001970197002.032.25014919770351002340437500
132014529802747.4634-121.987372772710271002.042.75013920000392992390630000
142014529802947.5794-122.025675182820282002.052.5013819790483512820675000
" + ], + "text/plain": [ + "+-----+------+-------+-----+---+----------------+---------------+------------------+--------+\n", + "| id | year | month | day | … | year_renovated | sqft_lot_15nn | sqft_living_15nn | price |\n", + "| --- | --- | --- | --- | | --- | --- | --- | --- |\n", + "| i64 | i64 | i64 | i64 | | i64 | i64 | i64 | i64 |\n", + "+===========================================================================================+\n", + "| 0 | 2014 | 5 | 2 | … | 0 | 9176 | 2310 | 285000 |\n", + "| 1 | 2014 | 5 | 2 | … | 0 | 8595 | 1750 | 285000 |\n", + "| 2 | 2014 | 5 | 2 | … | 0 | 9000 | 1850 | 440000 |\n", + "| 3 | 2014 | 5 | 2 | … | 0 | 8942 | 1260 | 435000 |\n", + "| 4 | 2014 | 5 | 2 | … | 0 | 10836 | 1450 | 430000 |\n", + "| … | … | … | … | … | … | … | … | … |\n", + "| 10 | 2014 | 5 | 2 | … | 0 | 24967 | 4050 | 604000 |\n", + "| 11 | 2014 | 5 | 2 | … | 0 | 16215 | 1450 | 335000 |\n", + "| 12 | 2014 | 5 | 2 | … | 0 | 35100 | 2340 | 437500 |\n", + "| 13 | 2014 | 5 | 2 | … | 0 | 39299 | 2390 | 630000 |\n", + "| 14 | 2014 | 5 | 2 | … | 0 | 48351 | 2820 | 675000 |\n", + "+-----+------+-------+-----+---+----------------+---------------+------------------+--------+" + ] + }, + "execution_count": 1, + "metadata": {}, + "output_type": "execute_result" + } + ], "source": [ "from safeds.data.tabular.containers import Table\n", "\n", @@ -57,9 +94,40 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 2, "metadata": {}, - "outputs": [], + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "shape: (5, 21)
idyearmonthdayzipcodesqft_lotsqft_livingsqft_abovesqft_basementfloorsbedroomsbathroomswaterfrontviewconditiongradeyear_builtyear_renovatedsqft_lot_15nnsqft_living_15nnprice
i64i64i64i64i64i64i64i64i64f64i64f64i64i64i64i64i64i64i64i64i64
02014529800193972200220002.042.501381987091762310285000
12014529800310834209013607301.032.501481987085951750285000
22014529800681192160108010801.042.2501381966090001850440000
32014529800688001450145001.041.001471954089421260435000
42014529800610000192010708501.041.5014719540108361450430000
" + ], + "text/plain": [ + "+-----+------+-------+-----+---+----------------+---------------+------------------+--------+\n", + "| id | year | month | day | … | year_renovated | sqft_lot_15nn | sqft_living_15nn | price |\n", + "| --- | --- | --- | --- | | --- | --- | --- | --- |\n", + "| i64 | i64 | i64 | i64 | | i64 | i64 | i64 | i64 |\n", + "+===========================================================================================+\n", + "| 0 | 2014 | 5 | 2 | … | 0 | 9176 | 2310 | 285000 |\n", + "| 1 | 2014 | 5 | 2 | … | 0 | 8595 | 1750 | 285000 |\n", + "| 2 | 2014 | 5 | 2 | … | 0 | 9000 | 1850 | 440000 |\n", + "| 3 | 2014 | 5 | 2 | … | 0 | 8942 | 1260 | 435000 |\n", + "| 4 | 2014 | 5 | 2 | … | 0 | 10836 | 1450 | 430000 |\n", + "+-----+------+-------+-----+---+----------------+---------------+------------------+--------+" + ] + }, + "execution_count": 2, + "metadata": {}, + "output_type": "execute_result" + } + ], "source": [ "pricing_columns = (\n", " # Removes columns \"latitude\" and \"longitude\" from table\n", @@ -76,7 +144,9 @@ { "cell_type": "markdown", "metadata": {}, - "source": "See how to perform further data cleaning in the dedicated [Data Processing Tutorial](../data_processing)." + "source": [ + "See how to perform further data cleaning in the dedicated [Data Processing Tutorial](../data_processing)." + ] }, { "cell_type": "markdown", @@ -91,7 +161,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 3, "metadata": { "collapsed": false }, @@ -111,7 +181,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 4, "metadata": { "collapsed": false }, @@ -135,7 +205,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 5, "metadata": { "collapsed": false }, @@ -159,11 +229,49 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 6, "metadata": { "collapsed": false }, - "outputs": [], + "outputs": [ + { + "data": { + "text/html": [ + "
\n", + "shape: (15, 21)
idyearmonthdayzipcodesqft_lotsqft_livingsqft_abovesqft_basementfloorsbedroomsbathroomswaterfrontviewconditiongradeyear_builtyear_renovatedsqft_lot_15nnsqft_living_15nnprice
i64i64i64i64i64i64i64i64i64f64i64f64i64i64i64i64i64i64i64i64f64
1054920141013981035000124010002401.021.001371920035001480550661.111111
1759020153149800273122010201001.042.001471976076502010269944.444444
10889201410179810332201120112001.021.001471923032201440550661.111111
12511201411149814424571950195003.022.501382009016391650334662.5
2057220154279805650381220122001.031.001561942050381140195130.555556
………………………………………………………
909720149189800374001130113001.041.001471969073791540223916.666667
845420149898023847084084001.031.00146196108450840109291.666667
7829201482698126402582082001.021.003561922057501410330190.0
19952201541798198101871120112001.031.7501371968087361900201880.0
1838220153269804259292210221002.042.501382004059012200311491.666667
" + ], + "text/plain": [ + "+-------+------+-------+-----+---+----------------+---------------+-----------------+--------------+\n", + "| id | year | month | day | … | year_renovated | sqft_lot_15nn | sqft_living_15n | price |\n", + "| --- | --- | --- | --- | | --- | --- | n | --- |\n", + "| i64 | i64 | i64 | i64 | | i64 | i64 | --- | f64 |\n", + "| | | | | | | | i64 | |\n", + "+==================================================================================================+\n", + "| 10549 | 2014 | 10 | 13 | … | 0 | 3500 | 1480 | 550661.11111 |\n", + "| 17590 | 2015 | 3 | 14 | … | 0 | 7650 | 2010 | 269944.44444 |\n", + "| 10889 | 2014 | 10 | 17 | … | 0 | 3220 | 1440 | 550661.11111 |\n", + "| 12511 | 2014 | 11 | 14 | … | 0 | 1639 | 1650 | 334662.50000 |\n", + "| 20572 | 2015 | 4 | 27 | … | 0 | 5038 | 1140 | 195130.55556 |\n", + "| … | … | … | … | … | … | … | … | … |\n", + "| 9097 | 2014 | 9 | 18 | … | 0 | 7379 | 1540 | 223916.66667 |\n", + "| 8454 | 2014 | 9 | 8 | … | 0 | 8450 | 840 | 109291.66667 |\n", + "| 7829 | 2014 | 8 | 26 | … | 0 | 5750 | 1410 | 330190.00000 |\n", + "| 19952 | 2015 | 4 | 17 | … | 0 | 8736 | 1900 | 201880.00000 |\n", + "| 18382 | 2015 | 3 | 26 | … | 0 | 5901 | 2200 | 311491.66667 |\n", + "+-------+------+-------+-----+---+----------------+---------------+-----------------+--------------+" + ] + }, + "execution_count": 6, + "metadata": {}, + "output_type": "execute_result" + } + ], "source": [ "prediction = fitted_model.predict(testing_table)\n", "# For visualisation purposes we only print out the first 15 rows.\n", @@ -183,11 +291,22 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 7, "metadata": { "collapsed": false }, - "outputs": [], + "outputs": [ + { + "data": { + "text/plain": [ + "92598.0054595909" + ] + }, + "execution_count": 7, + "metadata": {}, + "output_type": "execute_result" + } + ], "source": [ "fitted_model.mean_absolute_error(testing_table)" ] @@ -201,9 +320,20 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 8, "metadata": {}, - "outputs": [], + "outputs": [ + { + "data": { + "text/plain": [ + "92541.16556653795" + ] + }, + "execution_count": 8, + "metadata": {}, + "output_type": "execute_result" + } + ], "source": [ "from safeds.data.tabular.containers import Table\n", "from safeds.ml.classical.regression import DecisionTreeRegressor\n", diff --git a/src/safeds/ml/nn/_internal_model.py b/src/safeds/ml/nn/_internal_model.py index cb1105503..4c4f42d45 100644 --- a/src/safeds/ml/nn/_internal_model.py +++ b/src/safeds/ml/nn/_internal_model.py @@ -10,7 +10,7 @@ from safeds._config import _init_default_device from safeds.ml.nn.converters._input_converter_image import _InputConversionImage -from safeds.ml.nn.layers import FlattenLayer, Layer +from safeds.ml.nn.layers import DropoutLayer, FlattenLayer, Layer from safeds.ml.nn.layers._pooling2d_layer import _Pooling2DLayer if TYPE_CHECKING: @@ -36,7 +36,7 @@ def __init__(self, input_conversion: InputConversion, layers: list[Layer], is_fo layer._set_input_size(previous_output_size) elif isinstance(input_conversion, _InputConversionImage): layer._set_input_size(input_conversion._data_size) - if isinstance(layer, FlattenLayer | _Pooling2DLayer): + if isinstance(layer, FlattenLayer | _Pooling2DLayer | DropoutLayer): internal_layers.append(layer._get_internal_layer()) else: internal_layers.append(layer._get_internal_layer(activation_function="relu")) diff --git a/src/safeds/ml/nn/_model.py b/src/safeds/ml/nn/_model.py index 3d9086d61..aa2927af2 100644 --- a/src/safeds/ml/nn/_model.py +++ b/src/safeds/ml/nn/_model.py @@ -27,7 +27,7 @@ ForwardLayer, ) from safeds.ml.nn.layers._pooling2d_layer import _Pooling2DLayer -from safeds.ml.nn.typing import ConstantImageSize, ModelImageSize, VariableImageSize +from safeds.ml.nn.typing import ConstantImageSize, ModelImageSize, TensorShape, VariableImageSize if TYPE_CHECKING: from collections.abc import Callable @@ -108,6 +108,21 @@ def __init__( self._total_number_of_batches_done = 0 self._total_number_of_epochs_done = 0 + def get_parameter_count(self) -> int: + if self._input_size is None: + raise ValueError("The input_size is not yet set.") + + summand = 0 + last_type = "int" if isinstance(self.input_size, int) else "ImageSize" + last_input_neurons = self.input_size if isinstance(self.input_size, int) else 0 + last_input_channels = self.input_size.channel if isinstance(self.input_size, ModelImageSize) else 0 + for layer in self._layers: + layer._set_input_size(last_input_neurons if last_type == "int" else last_input_channels) + summand += layer.get_parameter_count(TensorShape([last_input_neurons, last_input_channels])) + last_input_neurons = layer.output_size if isinstance(layer.output_size, int) else 0 + last_input_channels = layer.output_size.channel if isinstance(layer.output_size, ModelImageSize) else 0 + return summand + @staticmethod def load_pretrained_model(huggingface_repo: str) -> NeuralNetworkRegressor: # pragma: no cover """ @@ -387,6 +402,21 @@ def __init__( self._total_number_of_batches_done = 0 self._total_number_of_epochs_done = 0 + def get_parameter_count(self) -> int: + if self._input_size is None: + raise ValueError("The input_size is not yet set.") + + summand = 0 + last_type = "int" if isinstance(self.input_size, int) else "ImageSize" + last_input_neurons = self.input_size if isinstance(self.input_size, int) else 0 + last_input_channels = self.input_size.channel if isinstance(self.input_size, ModelImageSize) else 0 + for layer in self._layers: + layer._set_input_size(last_input_neurons if last_type == "int" else last_input_channels) + summand += layer.get_parameter_count(TensorShape([last_input_neurons, last_input_channels])) + last_input_neurons = layer.output_size if isinstance(layer.output_size, int) else 0 + last_input_channels = layer.output_size.channel if isinstance(layer.output_size, ModelImageSize) else 0 + return summand + @staticmethod def load_pretrained_model(huggingface_repo: str) -> NeuralNetworkClassifier: # pragma: no cover """ diff --git a/src/safeds/ml/nn/layers/_convolutional2d_layer.py b/src/safeds/ml/nn/layers/_convolutional2d_layer.py index dd42f2d97..c59f02480 100644 --- a/src/safeds/ml/nn/layers/_convolutional2d_layer.py +++ b/src/safeds/ml/nn/layers/_convolutional2d_layer.py @@ -11,7 +11,7 @@ if TYPE_CHECKING: from torch import nn - from safeds.ml.nn.typing import ModelImageSize + from safeds.ml.nn.typing import ModelImageSize, TensorShape class Convolutional2DLayer(Layer): @@ -157,6 +157,9 @@ def __sizeof__(self) -> int: + sys.getsizeof(self._output_size) ) + def get_parameter_count(self, input_size: TensorShape) -> int: + return int((self._kernel_size * self._kernel_size * input_size._dims[1] + 1) * self._output_channel) + class ConvolutionalTranspose2DLayer(Convolutional2DLayer): """ @@ -261,3 +264,6 @@ def __eq__(self, other: object) -> bool: def __sizeof__(self) -> int: return sys.getsizeof(self._output_padding) + super().__sizeof__() + + def get_parameter_count(self, input_size: TensorShape) -> int: + return int((self._kernel_size * self._kernel_size * input_size._dims[1] + 1) * self._output_channel) diff --git a/src/safeds/ml/nn/layers/_dropout_layer.py b/src/safeds/ml/nn/layers/_dropout_layer.py index 1814e4383..af857ef90 100644 --- a/src/safeds/ml/nn/layers/_dropout_layer.py +++ b/src/safeds/ml/nn/layers/_dropout_layer.py @@ -4,7 +4,7 @@ from safeds._utils import _structural_hash from safeds._validation import _check_bounds, _ClosedBound -from safeds.ml.nn.typing import ModelImageSize +from safeds.ml.nn.typing import ModelImageSize, TensorShape from ._layer import Layer @@ -104,3 +104,6 @@ def __sizeof__(self) -> int: return int(self._input_size) elif isinstance(self._input_size, ModelImageSize): return self._input_size.__sizeof__() + + def get_parameter_count(self, input_size: TensorShape) -> int: # noqa: ARG002 + return 0 diff --git a/src/safeds/ml/nn/layers/_flatten_layer.py b/src/safeds/ml/nn/layers/_flatten_layer.py index a84551c2b..63a9e9cd2 100644 --- a/src/safeds/ml/nn/layers/_flatten_layer.py +++ b/src/safeds/ml/nn/layers/_flatten_layer.py @@ -4,7 +4,7 @@ from typing import TYPE_CHECKING, Any from safeds._utils import _structural_hash -from safeds.ml.nn.typing import ConstantImageSize +from safeds.ml.nn.typing import ConstantImageSize, TensorShape from ._layer import Layer @@ -86,3 +86,6 @@ def __eq__(self, other: object) -> bool: def __sizeof__(self) -> int: return sys.getsizeof(self._input_size) + sys.getsizeof(self._output_size) + + def get_parameter_count(self, input_size: TensorShape) -> int: # noqa: ARG002 + return 0 diff --git a/src/safeds/ml/nn/layers/_forward_layer.py b/src/safeds/ml/nn/layers/_forward_layer.py index e420b78ec..9f9335e5a 100644 --- a/src/safeds/ml/nn/layers/_forward_layer.py +++ b/src/safeds/ml/nn/layers/_forward_layer.py @@ -4,7 +4,7 @@ from safeds._utils import _structural_hash from safeds._validation import _check_bounds, _ClosedBound -from safeds.ml.nn.typing import ModelImageSize +from safeds.ml.nn.typing import ModelImageSize, TensorShape from ._layer import Layer @@ -96,3 +96,6 @@ def __sizeof__(self) -> int: import sys return sys.getsizeof(self._input_size) + sys.getsizeof(self._output_size) + + def get_parameter_count(self, input_size: TensorShape) -> int: + return (input_size._dims[0] + 1) * self._output_size diff --git a/src/safeds/ml/nn/layers/_gru_layer.py b/src/safeds/ml/nn/layers/_gru_layer.py index e74fec417..3ada7082b 100644 --- a/src/safeds/ml/nn/layers/_gru_layer.py +++ b/src/safeds/ml/nn/layers/_gru_layer.py @@ -5,7 +5,7 @@ from safeds._utils import _structural_hash from safeds._validation import _check_bounds, _ClosedBound -from safeds.ml.nn.typing import ModelImageSize +from safeds.ml.nn.typing import ModelImageSize, TensorShape from ._layer import Layer @@ -95,3 +95,6 @@ def __eq__(self, other: object) -> bool: def __sizeof__(self) -> int: return sys.getsizeof(self._input_size) + sys.getsizeof(self._output_size) + + def get_parameter_count(self, input_size: TensorShape) -> int: + return (input_size._dims[0] + self._output_size + 2) * self._output_size * 3 diff --git a/src/safeds/ml/nn/layers/_layer.py b/src/safeds/ml/nn/layers/_layer.py index 058036688..a95155da1 100644 --- a/src/safeds/ml/nn/layers/_layer.py +++ b/src/safeds/ml/nn/layers/_layer.py @@ -6,7 +6,7 @@ if TYPE_CHECKING: from torch import nn - from safeds.ml.nn.typing import ModelImageSize + from safeds.ml.nn.typing import ModelImageSize, TensorShape class Layer(ABC): @@ -43,3 +43,7 @@ def __eq__(self, other: object) -> bool: @abstractmethod def __sizeof__(self) -> int: pass # pragma: no cover + + @abstractmethod + def get_parameter_count(self, input_size: TensorShape) -> int: + pass # pragma: no cover diff --git a/src/safeds/ml/nn/layers/_lstm_layer.py b/src/safeds/ml/nn/layers/_lstm_layer.py index 330809474..06825d8af 100644 --- a/src/safeds/ml/nn/layers/_lstm_layer.py +++ b/src/safeds/ml/nn/layers/_lstm_layer.py @@ -5,7 +5,7 @@ from safeds._utils import _structural_hash from safeds._validation import _check_bounds, _ClosedBound -from safeds.ml.nn.typing import ModelImageSize +from safeds.ml.nn.typing import ModelImageSize, TensorShape from ._layer import Layer @@ -95,3 +95,6 @@ def __eq__(self, other: object) -> bool: def __sizeof__(self) -> int: return sys.getsizeof(self._input_size) + sys.getsizeof(self._output_size) + + def get_parameter_count(self, input_size: TensorShape) -> int: + return (input_size._dims[0] + self._output_size + 2) * self._output_size * 4 diff --git a/src/safeds/ml/nn/layers/_pooling2d_layer.py b/src/safeds/ml/nn/layers/_pooling2d_layer.py index d658ed848..7792b0044 100644 --- a/src/safeds/ml/nn/layers/_pooling2d_layer.py +++ b/src/safeds/ml/nn/layers/_pooling2d_layer.py @@ -11,7 +11,7 @@ if TYPE_CHECKING: from torch import nn - from safeds.ml.nn.typing import ModelImageSize + from safeds.ml.nn.typing import ModelImageSize, TensorShape class _Pooling2DLayer(Layer): @@ -134,6 +134,9 @@ def __sizeof__(self) -> int: + sys.getsizeof(self._padding) ) + def get_parameter_count(self, input_size: TensorShape) -> int: # noqa: ARG002 + return 0 + class MaxPooling2DLayer(_Pooling2DLayer): """ diff --git a/src/safeds/ml/nn/typing/__init__.py b/src/safeds/ml/nn/typing/__init__.py index 913ad6f65..f7cef6927 100644 --- a/src/safeds/ml/nn/typing/__init__.py +++ b/src/safeds/ml/nn/typing/__init__.py @@ -6,6 +6,7 @@ if TYPE_CHECKING: from ._model_image_size import ConstantImageSize, ModelImageSize, VariableImageSize + from ._tensor_shape import TensorShape apipkg.initpkg( __name__, @@ -13,6 +14,7 @@ "ConstantImageSize": "._model_image_size:ConstantImageSize", "ModelImageSize": "._model_image_size:ModelImageSize", "VariableImageSize": "._model_image_size:VariableImageSize", + "TensorShape": "._tensor_shape:TensorShape", }, ) @@ -20,4 +22,5 @@ "ConstantImageSize", "ModelImageSize", "VariableImageSize", + "TensorShape", ] diff --git a/src/safeds/ml/nn/typing/_tensor_shape.py b/src/safeds/ml/nn/typing/_tensor_shape.py new file mode 100644 index 000000000..f6c4b9d4f --- /dev/null +++ b/src/safeds/ml/nn/typing/_tensor_shape.py @@ -0,0 +1,59 @@ +from __future__ import annotations + +from safeds._utils import _structural_hash +from safeds._validation import _check_bounds, _ClosedBound + + +class TensorShape: + """ + Initializes a TensorShape object with the given dimensions. + + Parameters + ---------- + dims: + A list of integers where each integer represents + the size of the tensor in a particular dimension. + """ + + def __init__(self, dims: list[int]) -> None: + self._dims = dims + + def get_size(self, dimension: int | None = None) -> int: + """ + Return the size of the tensor in the specified dimension. + + Parameters. + ---------- + dimension: + The dimension index for which the size is to be retrieved. + + Returns + ------- + int: The size of the tensor in the specified dimension. + + Raises + ------ + OutOfBoundsError: + If the actual value is outside its expected range. + """ + _check_bounds("dimension", dimension, lower_bound=_ClosedBound(0)) + if dimension is not None and dimension >= self.dimensionality: + # TODO maybe add error message indicating that the dimension is out of range + return 0 + if dimension is None: + return self._dims[0] + return self._dims[dimension] + + def __hash__(self) -> int: + return _structural_hash(self._dims) + + @property + def dimensionality(self) -> int: + """ + Returns the number of dimensions of the tensor. + + Returns + ------- + int: The number of dimensions of the tensor. + """ + return len(self._dims) diff --git a/tests/safeds/ml/nn/layers/test_convolutional2d_layer.py b/tests/safeds/ml/nn/layers/test_convolutional2d_layer.py index 4b8ac362e..f79595919 100644 --- a/tests/safeds/ml/nn/layers/test_convolutional2d_layer.py +++ b/tests/safeds/ml/nn/layers/test_convolutional2d_layer.py @@ -4,6 +4,7 @@ import pytest from safeds.data.image.typing import ImageSize from safeds.ml.nn.layers import Convolutional2DLayer, ConvolutionalTranspose2DLayer +from safeds.ml.nn.typing import TensorShape from torch import nn @@ -157,6 +158,22 @@ def test_should_raise_if_input_size_is_set_with_int( with pytest.raises(TypeError, match=r"The input_size of a convolution layer has to be of type ImageSize."): layer._set_input_size(1) + def test_conv_get_parameter_count_returns_right_amount(self) -> None: + kernel_size = 5 + input_channels = 3 + output_channels = 3 + expected_output = int((kernel_size * kernel_size * input_channels + 1) * output_channels) + layer = Convolutional2DLayer(input_channels, kernel_size) + assert layer.get_parameter_count(TensorShape([1, input_channels])) == expected_output + + def test_conv_transposed_get_parameter_count_returns_right_amount(self) -> None: + kernel_size = 5 + input_channels = 3 + output_channels = 3 + expected_output = int((kernel_size * kernel_size * input_channels + 1) * output_channels) + layer = ConvolutionalTranspose2DLayer(input_channels, kernel_size) + assert layer.get_parameter_count(TensorShape([1, input_channels])) == expected_output + class TestEq: @pytest.mark.parametrize( ("conv2dlayer1", "conv2dlayer2"), diff --git a/tests/safeds/ml/nn/layers/test_dropout_layer.py b/tests/safeds/ml/nn/layers/test_dropout_layer.py index 5dad5f3a8..b9e47477b 100644 --- a/tests/safeds/ml/nn/layers/test_dropout_layer.py +++ b/tests/safeds/ml/nn/layers/test_dropout_layer.py @@ -4,7 +4,7 @@ from safeds.data.tabular.containers import Table from safeds.exceptions import OutOfBoundsError from safeds.ml.nn.layers import DropoutLayer -from safeds.ml.nn.typing import ConstantImageSize +from safeds.ml.nn.typing import ConstantImageSize, TensorShape from torch import nn @@ -43,6 +43,10 @@ def test_input_size_should_be_set(self) -> None: with pytest.raises(ValueError, match=r"The input_size is not yet set."): layer.__sizeof__() + def test_get_parameter_count_right_output(self) -> None: + layer = DropoutLayer(0.5) + assert layer.get_parameter_count(TensorShape([1])) == 0 + class TestEq: def test_should_be_equal(self) -> None: diff --git a/tests/safeds/ml/nn/layers/test_flatten_layer.py b/tests/safeds/ml/nn/layers/test_flatten_layer.py index 64db6127c..3adb15236 100644 --- a/tests/safeds/ml/nn/layers/test_flatten_layer.py +++ b/tests/safeds/ml/nn/layers/test_flatten_layer.py @@ -4,7 +4,7 @@ from safeds.data.image.typing import ImageSize from safeds.data.tabular.containers import Table from safeds.ml.nn.layers import FlattenLayer -from safeds.ml.nn.typing import VariableImageSize +from safeds.ml.nn.typing import TensorShape, VariableImageSize from torch import nn @@ -37,6 +37,10 @@ def test_should_raise_if_input_size_is_set_with_variable_image_size(self) -> Non with pytest.raises(TypeError, match=r"The input_size of a flatten layer has to be a ConstantImageSize."): layer._set_input_size(VariableImageSize(1, 2, 3)) + def test_get_parameter_count_right_output(self) -> None: + layer = FlattenLayer() + assert layer.get_parameter_count(TensorShape([1])) == 0 + class TestEq: def test_should_be_equal(self) -> None: assert FlattenLayer() == FlattenLayer() diff --git a/tests/safeds/ml/nn/layers/test_forward_layer.py b/tests/safeds/ml/nn/layers/test_forward_layer.py index 0ecd3bd05..1b250f5b7 100644 --- a/tests/safeds/ml/nn/layers/test_forward_layer.py +++ b/tests/safeds/ml/nn/layers/test_forward_layer.py @@ -5,6 +5,7 @@ from safeds.data.image.typing import ImageSize from safeds.exceptions import OutOfBoundsError from safeds.ml.nn.layers import ForwardLayer +from safeds.ml.nn.typing import TensorShape from torch import nn # TODO: Should be tested on a model, not a layer, since input size gets inferred @@ -177,3 +178,11 @@ def test_should_assert_that_different_forward_layers_have_different_hash( ) def test_should_assert_that_layer_size_is_greater_than_normal_object(layer: ForwardLayer) -> None: assert sys.getsizeof(layer) > sys.getsizeof(object()) + + +def test_conv_transposed_get_parameter_count_returns_right_amount() -> None: + input_neurons = 3 + output_neurons = 3 + expected_output = int((input_neurons + 1) * output_neurons) + layer = ForwardLayer(output_neurons) + assert layer.get_parameter_count(TensorShape([input_neurons])) == expected_output diff --git a/tests/safeds/ml/nn/layers/test_gru_layer.py b/tests/safeds/ml/nn/layers/test_gru_layer.py index 4a6f366e4..930805771 100644 --- a/tests/safeds/ml/nn/layers/test_gru_layer.py +++ b/tests/safeds/ml/nn/layers/test_gru_layer.py @@ -5,6 +5,7 @@ from safeds.data.image.typing import ImageSize from safeds.exceptions import OutOfBoundsError from safeds.ml.nn.layers import GRULayer +from safeds.ml.nn.typing import TensorShape from torch import nn @@ -187,3 +188,11 @@ def test_internal_layer_should_raise_error() -> None: layer = GRULayer(1) with pytest.raises(ValueError, match="The input_size is not yet set."): layer._get_internal_layer(activation_function="relu") + + +def test_conv_transposed_get_parameter_count_returns_right_amount() -> None: + input_neurons = 4 + output_neurons = 16 + expected_output = int((input_neurons + output_neurons + 2) * output_neurons * 3) + layer = GRULayer(output_neurons) + assert layer.get_parameter_count(TensorShape([input_neurons])) == expected_output diff --git a/tests/safeds/ml/nn/layers/test_lstm_layer.py b/tests/safeds/ml/nn/layers/test_lstm_layer.py index 8d58e5dd8..2230d1839 100644 --- a/tests/safeds/ml/nn/layers/test_lstm_layer.py +++ b/tests/safeds/ml/nn/layers/test_lstm_layer.py @@ -5,6 +5,7 @@ from safeds.data.image.typing import ImageSize from safeds.exceptions import OutOfBoundsError from safeds.ml.nn.layers import LSTMLayer +from safeds.ml.nn.typing import TensorShape from torch import nn # TODO: Should be tested on a model, not a layer, since input size gets inferred @@ -177,3 +178,11 @@ def test_should_assert_that_different_forward_layers_have_different_hash( ) def test_should_assert_that_layer_size_is_greater_than_normal_object(layer: LSTMLayer) -> None: assert sys.getsizeof(layer) > sys.getsizeof(object()) + + +def test_conv_transposed_get_parameter_count_returns_right_amount() -> None: + input_neurons = 4 + output_neurons = 16 + expected_output = int((input_neurons + output_neurons + 2) * output_neurons * 4) + layer = LSTMLayer(output_neurons) + assert layer.get_parameter_count(TensorShape([input_neurons])) == expected_output diff --git a/tests/safeds/ml/nn/layers/test_pooling2d_layer.py b/tests/safeds/ml/nn/layers/test_pooling2d_layer.py index 4c5fcb6e3..e50d34221 100644 --- a/tests/safeds/ml/nn/layers/test_pooling2d_layer.py +++ b/tests/safeds/ml/nn/layers/test_pooling2d_layer.py @@ -6,6 +6,7 @@ from safeds.data.tabular.containers import Table from safeds.ml.nn.layers import AveragePooling2DLayer, MaxPooling2DLayer from safeds.ml.nn.layers._pooling2d_layer import _Pooling2DLayer +from safeds.ml.nn.typing import TensorShape from torch import nn @@ -56,6 +57,17 @@ def test_should_raise_if_input_size_is_set_with_int(self, strategy: Literal["max with pytest.raises(TypeError, match=r"The input_size of a pooling layer has to be of type ImageSize."): layer._set_input_size(1) + @pytest.mark.parametrize( + "strategy", + [ + "max", + "avg", + ], + ) + def test_get_parameter_count_right_output(self, strategy: Literal["max", "avg"]) -> None: + layer = _Pooling2DLayer(strategy, 2, stride=2, padding=2) + assert layer.get_parameter_count(TensorShape([1])) == 0 + class TestEq: @pytest.mark.parametrize( ("pooling_2d_layer_1", "pooling_2d_layer_2"), diff --git a/tests/safeds/ml/nn/test_model.py b/tests/safeds/ml/nn/test_model.py index 43fc67aa6..d14358e3f 100644 --- a/tests/safeds/ml/nn/test_model.py +++ b/tests/safeds/ml/nn/test_model.py @@ -27,6 +27,7 @@ AveragePooling2DLayer, Convolutional2DLayer, ConvolutionalTranspose2DLayer, + DropoutLayer, FlattenLayer, ForwardLayer, Layer, @@ -514,6 +515,27 @@ def test_should_be_pickleable(self, device: Device) -> None: # Should not raise pickle.dumps(fitted_model) + def test_parameters_model_not_fitted(self, device: Device) -> None: + configure_test_with_device(device) + model = NeuralNetworkClassifier( + InputConversionTable(), + [ForwardLayer(neuron_count=8), DropoutLayer(0.5), ForwardLayer(neuron_count=1)], + ) + with pytest.raises(ValueError, match=r"The input_size is not yet set."): + model.get_parameter_count() + + def test_should_sum_parameters(self, device: Device) -> None: + configure_test_with_device(device) + expected_output = 16 + 0 + 9 + model_fitted = NeuralNetworkClassifier( + InputConversionTable(), + [ForwardLayer(neuron_count=8), DropoutLayer(0.5), ForwardLayer(neuron_count=1)], + ).fit( + Table.from_dict({"a": [1, 1, 1], "b": [2, 2, 2]}).to_tabular_dataset("a"), + epoch_size=3, + ) + assert expected_output == model_fitted.get_parameter_count() + @pytest.mark.parametrize("device", get_devices(), ids=get_devices_ids()) class TestRegressionModel: @@ -879,3 +901,24 @@ def test_should_be_pickleable(self, device: Device) -> None: # Should not raise pickle.dumps(fitted_model) + + def test_parameters_model_not_fitted(self, device: Device) -> None: + configure_test_with_device(device) + model = NeuralNetworkRegressor( + InputConversionTable(), + [ForwardLayer(neuron_count=8), DropoutLayer(0.5), ForwardLayer(neuron_count=1)], + ) + with pytest.raises(ValueError, match=r"The input_size is not yet set."): + model.get_parameter_count() + + def test_should_sum_parameters(self, device: Device) -> None: + configure_test_with_device(device) + expected_output = 16 + 0 + 9 + model_fitted = NeuralNetworkRegressor( + InputConversionTable(), + [ForwardLayer(neuron_count=8), DropoutLayer(0.5), ForwardLayer(neuron_count=1)], + ).fit( + Table.from_dict({"a": [1, 1, 1], "b": [2, 2, 2]}).to_tabular_dataset("a"), + epoch_size=3, + ) + assert expected_output == model_fitted.get_parameter_count()