torchling 0.1.18

A minimal deep learning library using numpy.

torchling

A minimal neural network library. torchling is a work in progress.

Installation

pip install torchling

Example: Boston Housing Regression

import numpy as np
import pandas as pd
from sklearn.preprocessing import StandardScaler

from torchling.models import Sequential
from torchling.layers import Linear, Tanh
from torchling.losses import MSE
from torchling.optimisers import SGD


def load_dataset():
    url = "https://archive.ics.uci.edu/ml/machine-learning-databases/housing/housing.data"
    df = pd.read_csv(url, sep=r'\s+')
    X = df.iloc[:, :-1].values.astype(np.float64)
    Y = df.iloc[:, -1].values.astype(np.float64)
    return X, Y


def normalize(X, Y):
    scaler_X = StandardScaler()
    scaler_Y = StandardScaler()
    X_norm = scaler_X.fit_transform(X)
    Y_norm = scaler_Y.fit_transform(Y.reshape(-1, 1)).flatten()
    return X_norm, Y_norm, scaler_X, scaler_Y


def denormalize(scaler, arr):
    return scaler.inverse_transform(np.array(arr).reshape(-1, 1)).flatten()


X, Y = load_dataset()
X, Y, _, Y_metadata = normalize(X, Y)

X_TRAIN, Y_TRAIN = X[1:], Y[1:]
X_TEST, Y_TEST = X[:1], Y[0]

epochs = 3000
lr = 0.6
input_size = X_TRAIN.shape[1]

model = Sequential(
    Linear(50, input_size=input_size),
    Tanh(),
    Linear(50),
    Tanh(),
    Linear(1),
)

optimizer = SGD(alpha=lr)
loss = MSE()
model.train(X_TRAIN, Y_TRAIN, optimizer, loss, epochs=epochs, print_every=250)

pred = model.predict(X_TEST)
print(f"Prediction - {denormalize(Y_metadata, pred)} | Actual - {denormalize(Y_metadata, [Y_TEST])}")

License

MIT