Standardize

class myoverse.transforms.Standardize(mean, std, eps=1e-08, **kwargs)

Standardize using pre-computed mean and std.

Useful when you have statistics from the training set.

Parameters:

• mean (float | torch.Tensor) – Mean value(s) to subtract.

• std (float | torch.Tensor) – Standard deviation(s) to divide by.

• eps (float) – Small value to avoid division by zero.

Examples

>>> # Compute stats on training data
>>> train_mean = train_data.mean()
>>> train_std = train_data.std()
>>>
>>> # Apply to test data
>>> standardize = Standardize(mean=train_mean, std=train_std)
>>> test_normalized = standardize(test_data)

Methods

__init__(mean, std[, eps])
_apply(x)	Apply the transform.