Simulates trajectories based on u0 and f. Trajectory i is based on u0[i, :] and f[i, :, :] Args: u0: (N,s), N is the number of s

Function: get save name

input: [N, more than 2*(1+3), padded, padded]

upsample the wavelet coefficients to the target shape (old) w_sub: [N, layer, nt(maybe padded), nx(maybe padded)] (old) w: [N, layer, sha

Function: Generate densitys or zero_densitys gif saved at gif_dir Input: densitys: numpy array [256,128,128]

low, init: rescaled init_u: not rescaled

Internal function to register the TensorFlow backend. This function is called automatically once a TFSimulation is instantiated. :return: True if

Builds a sparse matrix such that when applied to a flattened pressure field, it calculates the laplace of that field, taking into account obstacles a

upsample the wavelet coefficients to the target shape w_sub: [N, nt, layer, nx, nx] tyoe: 'time' or 'space' w: [N, 2*nt, layer, nx, n

Function: write velocity density field for turns

return: repeat Yh[i] 2**i times if pad: [Yl.shape[0], Yl.shape[1], 1+3*J, 64, 64] else: [Yl.shape[0], Yl.shape[1], 1+3*J, Yh[0].shape[-2

Solve the linear system of equations Ax=k using the conjugate gradient (CG) algorithm. The implementation is based on https://nvlpubs.nist.gov/nistpu

Arguments: u_target: (batch_size, Nt, Nx) u: (batch_size, Nt, Nx) f: (batch_size, Nt - 1, Nx)

Samples from diffusion model and evaluates results. Args: ddpm: Diffusion model args: Configuration arguments cu