Pyray2 Deep Learning Framework

Pyray2 is a deep learning framework implemnted from scratch using python and libraries such as numpy, matplotlib, pandas.

How to Install

 $ pip install pyray2

How to use ?

Check Colab:

• Train on MNIST with Linear Layers
• Train on MNIST with CNN
• Lenet archeticture

UML Design

Supported Features

Layers

• Convolution 2D
• Dense
• Max Pooling
• Avg Pooling

Activations

• Sigmoid
• RelU
• Softmax
• Tanh

Losses

• Mean Square Error
• Cross Entropy

How to define a Model:

# Example
model = Model()
model.add(conv(1, 4, 3, padding=1))
model.add(MaxPool2D(kernel_size=(2,2)))
model.add(ReLU())
model.add(Flatten())
model.add(Dense(8*7*7, 10))
model.set_loss(CrossEntropyLoss())

Optimizers

• Gradiant Decent
• Momentum
• Adam

optim = GradientDecent(model.parameters(), learning_rate = 0.01)
optim = MomentumGD(model.parameters(), learning_rate = 0.01)
optim = Adam(model.parameters(), learning_rate = 0.01)
lr_schedular = StepLR(optimizer, step_size = 1, gamma=0.1)

Visualization

• View dataset's examples
• Plotting for loss
• live graph

# Show Dataset Examples
image, labels = dataloader.__getitem__(1)
images = image.T.reshape(batch_size, 1, width, height)
images = np.asarray(images)
img_viewer_examples(images, labels.tolist()[0], greyscale= True)

# Plotting Loss
model.graph()
# Training Loop ...

# Before the loop to start new thread
model.startGraph()

# Training Loop ...

# After the loop to close the thread
model.stopGraph()

Evaluation

• Precision
• Recall
• F1 Score

e = Evaluation(10)
for image, label in dataloader_test:
    predicted = .....
    pred = .....
    e.add_prediction(pred[np.newaxis],label)
print("the confusion Matrix :",e.get_confusion_Matrix())
print("the Mean F1 Score =",e.evaluate())

Dataset & Data Loader

• CSV dataset loading
• Batch loading by Data Loader
• Interface for MNIST & CIFAR-10 datasets
• Simillar to Pytorch Interface.

batch_size = 4
dataset = MNIST_dataset("train.csv")
dataloader = Data_Loader(dataset, batch_size)
for image, label in dataloader:
	...
	...

Save / Load utils

• Saving / Loading model weights through .pth files.

model = load_weights(path)
save_weights(model, path)

Architectures

• LeNet
• AlexNet