Initialize bias the same way as torch.nn.Linear.

Initialize bias the same way as torch.nn.Linear.

Initialize bias the same way as torch.nn.Linear.

Return soft permutation of shape sample_shape + (size, size)

Multiply \sum_i Krylov(Z_f, v_i) @ w_i. Parameters: v: (nstack, rank, n) w: (batch_size, nstack, rank, n) f: real number

Multiply Krylov(Z_f, v_i)^T @ u. Parameters: v: (nstack, rank, n) u: (batch_size, n) f: real number Returns: p

The indices in a n x n matrix that marks where the entries of a butterfly factors are.

Return the bit reversal permutation used in discrete wavelet transform. Example: [0, 1, ..., 7] -> [0, 4, 2, 6, 1, 3, 5, 7] By default, the pe

In-place multiply the whole butterfly matrix by some scale factor, by multiplying the twiddle. Scale must be nonnegative

In-place multiply the whole butterfly matrix by some scale factor, by multiplying the twiddle. Scale must be nonnegative

Multiply by diagonal matrix Parameter: size: int diagonal_init: (n, )

Fixed permutation. Parameter: permutation: (n, ) tensor of ints

Perform map1 on the last dimension of the input and then map2 on the next to last dimension.

Parameters: diagonal_init: (out_channels, in_channels, size)

writes both to stream and to file. file_filters is a list of callables that processes a string just before being written to t