example.py

from DeepStorm.model import Model
from DeepStorm.Layers.conv import Conv2d
from DeepStorm.Layers.batch_normalization import BatchNorm2d
from DeepStorm.Layers.dropout import Dropout
from DeepStorm.Layers.pooling import MaxPool2d
from DeepStorm.Layers.flatten import Flatten
from DeepStorm.Layers.linear import Linear
from DeepStorm.Initializers.xavier import Xavier
from DeepStorm.Initializers.he import He
from DeepStorm.Initializers.uniform_random import UniformRandom
from DeepStorm.Initializers.constant import Constant
from DeepStorm.Activations.relu import ReLU
from DeepStorm.Activations.sigmoid import Sigmoid
from DeepStorm.Activations.softmax import SoftMax
from DeepStorm.Optimizers.adam import Adam
from DeepStorm.Losses.cross_entropy import CrossEntropyLoss

import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
from sklearn.model_selection import train_test_split

input_folder_path = "Data/"
train_df = pd.read_csv(f"{input_folder_path}train.csv")[:1000]
test_df = pd.read_csv(f"{input_folder_path}test.csv")

train_labels = train_df['label'].values
train_images = (train_df.iloc[:, 1:].values).astype('float32')
test_images = (test_df.iloc[:, :].values).astype('float32')

train_images, val_images, train_labels, val_labels = train_test_split(train_images, train_labels,
                                                                      stratify=None, random_state=123,
                                                                      test_size=0.20)
train_images = train_images.reshape(train_images.shape[0], 28 * 28)
val_images = val_images.reshape(val_images.shape[0], 28 * 28)
test_images = test_images.reshape(test_images.shape[0], 28 * 28)

train_images = train_images/255.0
val_images = val_images/255.0
test_images = test_images/255.0

classes = 10

train_labels = train_labels.reshape(-1)
train_labels = np.eye(classes)[train_labels]
val_labels = val_labels.reshape(-1)
val_labels = np.eye(classes)[val_labels]

layers = [
    Linear(in_features=28*28, out_features=256),
    ReLU(),
    Linear(in_features=256, out_features=128),
    ReLU(),
    Linear(in_features=128, out_features=64),
    ReLU(),
    Linear(in_features=64, out_features=10),
    SoftMax(),
]

model = Model(layers)

batch_size = 64
model.compile(optimizer=Adam(learning_rate=5e-3, mu=0.98, rho=0.999), loss=CrossEntropyLoss(),
              batch_size=batch_size, metrics=['accuracy'])

epochs = 20
history = model.fit(x_train=train_images, y_train=train_labels,
                    x_val=val_images, y_val=val_labels, epochs=epochs)