Hands-On Machine Learning for Algorithmic Trading

How to create a PyTorch DataLoader

We begin by converting the NumPy or pandas input data to Torch tensors. Conversion from and to NumPy is very straightforward:

import torchX_tensor = torch.from_numpy(X)y_tensor = torch.from_numpy(y)X_tensor.shape, y_tensor.shape(torch.Size([50000, 2]), torch.Size([50000]))

We can use these PyTorch tensors to instantiate first a TensorDataset instance and, in a second step, a DataLoader that includes information about batch_size:

import torch.utils.data as utilsdataset = utils.TensorDataset(X_tensor,y_tensor)dataloader = utils.DataLoader(dataset,                              batch_size=batch_size,                              shuffle=True)