ptycho

ptycho#

Functions for ptychography.

reconstruct

Solve the ptychography problem.

reconstruct_multigrid

Solve the ptychography problem using a multi-grid method.

Reconstruction

Context manager for online ptychography reconstruction.

simulate

Return real-valued detector counts of simulated ptychography data.
class tike.ptycho.ptycho.Reconstruction[source]#

Bases: object

Context manager for online ptychography reconstruction.

New in version 0.22.0.

Uses same parameters as the functional reconstruct API. Using a context manager allows for getting the current result or adding additional data without deallocating memory on the GPU.

Example

Data structure remain allocated on the GPU as long as the context is active. This allows quickly resuming a reconstruction.

with tike.ptycho.Reconstruction(
    data,
    parameters,
    num_gpu,
    use_mpi,
) as context:
    # Start reconstructing
    context.iterate(num_iter)
    # Check the current solution for the object
    early_result = context.get_psi()
    # Continue reconstructing
    context.iterate()
# All datastructures are transferred off the GPU at context close
final_result = context.parameters

See also

tike.ptycho.ptycho.reconstruct()

__init__(data, parameters, num_gpu=1, use_mpi=False)[source]#
Parameters:

  • data (numpy.typing.NDArray) –

  • parameters (PtychoParameters) –

  • num_gpu (Union[int, Tuple[int, ...]]) –

  • use_mpi (bool) –

append_new_data(new_data, new_scan)[source]#

Append new diffraction patterns and positions to existing result.

Parameters:

  • new_data (numpy.typing.NDArray) –

  • new_scan (numpy.typing.NDArray) –

Return type:

None

get_convergence()[source]#

Return the cost function values and times as a tuple.

Return type:

Tuple[List[List[float]], List[float]]

get_probe()[source]#

Return the current probe, eigen_probe, weights as numpy arrays.

Return type:

Tuple[numpy.array, numpy.array, numpy.array]

get_psi()[source]#

Return the current object estimate as a numpy array.

Return type:

numpy.array

get_result()[source]#

Return the current parameter estimates.

Return type:

PtychoParameters

get_scan()[source]#
Return type:

numpy.typing.NDArray

iterate(num_iter)[source]#

Advance the reconstruction by num_iter epochs.

Parameters:

num_iter (int) –

Return type:

None

tike.ptycho.ptycho.reconstruct(data, parameters, num_gpu=1, use_mpi=False)[source]#

Solve the ptychography problem.

This functional interface is the simplest to use, but deallocates GPU memory when it returns. Use the context manager API for the ability to get live results without ending the reconsturction.

Parameters:

  • data ((FRAME, WIDE, HIGH) uint16) – The intensity (square of the absolute value) of the propagated wavefront; i.e. what the detector records. FFT-shifted so the diffraction peak is at the corners.

  • parameters (tike.ptycho.solvers.PtychoParameters) – A class containing reconstruction parameters.

  • num_gpu (int, tuple(int)) – The number of GPUs to use or a tuple of the device numbers of the GPUs to use. If the number of GPUs is less than the requested number, only workers for the available GPUs are allocated.

  • use_mpi (bool) – Whether to use MPI or not.

Return type:

PtychoParameters

Changed in version 0.25.0: Removed the model parameter. Use tike.ptycho.exitwave.ExitwaveOptions instead.

Raises:

ValueError – When shapes are incorrect for various input parameters.

Returns:

result – A class containing reconstruction parameters.

Return type:

tike.ptycho.solvers.PtychoParameters

Parameters:

  • data (numpy.typing.NDArray) –

  • parameters (PtychoParameters) –

  • num_gpu (Union[int, Tuple[int, ...]]) –

  • use_mpi (bool) –

See also

tike.ptycho.ptycho.reconstruct_multigrid(), tike.ptycho.ptycho.Reconstruction()

tike.ptycho.ptycho.reconstruct_multigrid(data, parameters, num_gpu=1, use_mpi=False, num_levels=3, interp=<function _resize_fft>)[source]#

Solve the ptychography problem using a multi-grid method.

New in version 0.23.2.

Uses the same parameters as the functional reconstruct API. This function applies a multi-grid approach to the problem by downsampling the real-space input parameters and cropping the diffraction patterns to reduce the computational cost of early iterations.

Parameters:

  • num_levels (int > 0) – The number of times to reduce the problem by a factor of two.

  • data (numpy.typing.NDArray) –

  • parameters (PtychoParameters) –

  • num_gpu (Union[int, Tuple[int, ...]]) –

  • use_mpi (bool) –

  • interp (Callable) –

Return type:

PtychoParameters

See also

tike.ptycho.ptycho.reconstruct()

tike.ptycho.ptycho.simulate(detector_shape, probe, scan, psi, fly=1, eigen_probe=None, eigen_weights=None, **kwargs)[source]#

Return real-valued detector counts of simulated ptychography data.

Parameters:

  • detector_shape (int) – The pixel width of the detector.

  • probe ((..., 1, 1, SHARED, WIDE, HIGH) complex64) – The shared complex illumination function amongst all positions.

  • scan ((..., POSI, 2) float32) – Coordinates of the minimum corner of the probe grid for each measurement in the coordinate system of psi.

  • psi ((..., WIDE, HIGH) complex64) – The complex wavefront modulation of the object.

  • fly (int) – The number of scan positions which combine for one detector frame.

  • eigen_probe ((..., 1, EIGEN, SHARED, WIDE, HIGH) complex64) – The eigen probes for all positions.

  • eigen_weights ((..., POSI, EIGEN, SHARED) float32) – The relative intensity of the eigen probes at each position.

Returns:

data – The simulated intensity on the detector.

Return type:

(…, FRAME, WIDE, HIGH) float32

Contents