Utilities

Data

class pyroved.utils.init_dataloader(*args, random_sampler=False, shuffle=True, **kwargs)

Returns initialized PyTorch dataloader, which is used by pyroVED’s trainers. The inputs are torch Tensor objects containing training data and (optionally) labels.

Example:

>>> # Load training data stored as numpy array
>>> train_data = np.load("my_training_data.npy")
>>> # Transform numpy array to toech Tensor object
>>> train_data = torch.from_numpy(train_data).float()
>>> # Initialize dataloader
>>> train_loader = init_dataloader(train_data)

Return type

Type[DataLoader]

args = None

kwargs = None

random_sampler = None

return = None

shuffle = None

class pyroved.utils.init_ssvae_dataloaders(data_unsup, data_sup, data_val, **kwargs)

Helper function to initialize dataloader for ss-VAE models

Return type

Tuple[Type[DataLoader]]

data_sup = None

data_unsup = None

data_val = None

kwargs = None

return = None

Coordinates

class pyroved.utils.generate_grid(data_dim)

Generates 1D or 2D grid of coordinates

Return type

Tensor

data_dim = None

return = None

class pyroved.utils.generate_latent_grid(d, **kwargs)

Generates a grid of latent space coordinates

Return type

Tensor

d = None

return = None

class pyroved.utils.transform_coordinates(coord, phi=0, coord_dx=0)

Rotation of coordinates followed by translation. For 1D grid, there is only transaltion. Operates on batches.

Return type

Tensor

coord = None

coord_dx = None

phi = None

return = None

Visualization

class pyroved.utils.plot_img_grid(imgdata, d, **kwargs)

Plots a d-by-d square grid of 2D images

Return type

None

d = None

imgdata = None

kwargs = None

return = None

class pyroved.utils.plot_spect_grid(spectra, d, **kwargs)

Plots a d-by-d square grid with 1D spectral plots

d = None

kwargs = None

spectra = None