From a6b1df50f2ad8f6fe5437827a161a7810b3bb881 Mon Sep 17 00:00:00 2001
From: Jonathan Weine <jweine@ethz.ch>
Date: Mon, 31 Jul 2023 16:20:53 +0200
Subject: [PATCH 1/3] Added jit-compilations to contrast modules
---
.../analytic/contrast/coil_sensitivities.py | 5 ++---
.../analytic/contrast/diffusion_weighting.py | 7 ++++---
cmrsim/analytic/contrast/offresonance.py | 14 +++++++------
cmrsim/analytic/contrast/phase_tracking.py | 20 ++++++++++++-------
cmrsim/analytic/contrast/sequences.py | 11 +++++-----
cmrsim/analytic/contrast/t2_star.py | 9 +++++----
6 files changed, 38 insertions(+), 28 deletions(-)
diff --git a/cmrsim/analytic/contrast/coil_sensitivities.py b/cmrsim/analytic/contrast/coil_sensitivities.py
index 162bf1f..5bc3757 100644
--- a/cmrsim/analytic/contrast/coil_sensitivities.py
+++ b/cmrsim/analytic/contrast/coil_sensitivities.py
@@ -33,7 +33,6 @@ class CoilSensitivity(LookUpTableModule):
"""
required_quantities = ('r_vectors', )
- _device2: str
#: Spatial extend of the look up as: (3, 2) -> [(xlow, xhigh), (ylow, yhigh), (zlow, zhigh)]
map_dimensions: tf.Variable # float32
@@ -61,7 +60,6 @@ class CoilSensitivity(LookUpTableModule):
name="coil_sensitivities",
device=device,
expand_repetitions=True)
- self._device2 = device
self.expansion_factor = coils.shape[-1] // 2
def update(self):
@@ -69,6 +67,7 @@ class CoilSensitivity(LookUpTableModule):
self.expansion_factor = self.look_up_map3d.read_value().shape[-1] // 2
# pylint: disable=signature-differs, too-many-locals
+ @tf.function(jit_compile=True)
def __call__(self, signal_tensor: tf.Tensor, r_vectors: tf.Tensor, **kwargs) -> tf.Tensor:
"""Evaluation of coil-sensitivity weighting:
@@ -80,7 +79,7 @@ class CoilSensitivity(LookUpTableModule):
:param r_vectors: (#voxels, #repetitions, #samples, 3)
:return:
"""
- with tf.device(self._device2):
+ with tf.device(self.device):
input_shape = tf.shape(signal_tensor)
tf.Assert(tf.shape(r_vectors)[1] == 1 or
tf.shape(r_vectors)[1] == input_shape[1],
diff --git a/cmrsim/analytic/contrast/diffusion_weighting.py b/cmrsim/analytic/contrast/diffusion_weighting.py
index 0ce105b..16c4455 100644
--- a/cmrsim/analytic/contrast/diffusion_weighting.py
+++ b/cmrsim/analytic/contrast/diffusion_weighting.py
@@ -33,10 +33,11 @@ class GaussianDiffusionTensor(BaseSignalModel):
:param expand_repetitions: See BaseSignalModel for explanation
"""
super().__init__(name="dti_gauss", expand_repetitions=expand_repetitions, **kwargs)
- self.b_vectors = tf.Variable(b_vectors, shape=(None, 3), dtype=tf.float32, name='b_vectors')
- self.update()
+ with tf.device(self.device):
+ self.b_vectors = tf.Variable(b_vectors, shape=(None, 3), dtype=tf.float32, name='b_vectors')
+ self.update()
- @tf.Module.with_name_scope
+ @tf.function(jit_compile=True)
def __call__(self, signal_tensor: tf.Tensor, diffusion_tensor: tf.Tensor, **kwargs)\
-> tf.Tensor:
""" Evaluates the diffusion operator.
diff --git a/cmrsim/analytic/contrast/offresonance.py b/cmrsim/analytic/contrast/offresonance.py
index 2c8d9d4..67cd5a4 100644
--- a/cmrsim/analytic/contrast/offresonance.py
+++ b/cmrsim/analytic/contrast/offresonance.py
@@ -15,7 +15,7 @@ class OffResonanceProperty(BaseSignalModel):
.. math::
- M_{xy}^{i, \\prime} = M_{xy}^{i} exp(-\gamma \\Delta B \\delta t),
+ M_{xy}^{i, \\prime} = M_{xy}^{i} exp(-j\gamma \\Delta B \\delta t),
where :math:`\\delta t` is the duration from the RF-excitation center for GRE sequences
while it is the temporal distance to the echo-center for SE sequences.
@@ -53,17 +53,19 @@ class OffResonanceProperty(BaseSignalModel):
f"(#repetitions, #k-space-points), but is: {sampling_times.shape}")
if len(sampling_times.shape) == 1:
sampling_times = tf.reshape(sampling_times, [1, -1])
-
- self.gamma = tf.constant(gamma, dtype=tf.float32)
- self.sampling_times = tf.Variable(tf.constant(sampling_times, tf.float32),
- shape=(None, None), dtype=tf.float32)
- self.expansion_factor = sampling_times.shape[0]
+
+ with tf.device(self.device):
+ self.gamma = tf.constant(gamma, dtype=tf.float32)
+ self.sampling_times = tf.Variable(tf.constant(sampling_times, tf.float32),
+ shape=(None, None), dtype=tf.float32)
+ self.expansion_factor = sampling_times.shape[0]
def update(self):
super().update()
self.expansion_factor == tf.shape(self.sampling_times.read_value())[0]
# pylint: disable=signature-differs
+ @tf.function(jit_compile=True)
def __call__(self, signal_tensor: tf.Tensor, deltaB0: tf.Tensor, **kwargs):
""" Evaluates the off-resonance operator under the assumption
diff --git a/cmrsim/analytic/contrast/phase_tracking.py b/cmrsim/analytic/contrast/phase_tracking.py
index 81b704a..118aec3 100644
--- a/cmrsim/analytic/contrast/phase_tracking.py
+++ b/cmrsim/analytic/contrast/phase_tracking.py
@@ -22,6 +22,10 @@ class PhaseTracking(BaseSignalModel):
:param gamma: Gyromagnetic ratio in rad/ms/mT
"""
required_quantities = ('trajectory', )
+ #: (#repetitions, #time-steps, 4[t, x, y, z])
+ wave_form: tf.Variable
+ #: gyromagnetic ratio in rad/mT/ms
+ gamma: tf.Tensor
def __init__(self, expand_repetitions: bool, wave_form: tf.Tensor,
gamma: float, **kwargs):
@@ -31,13 +35,15 @@ class PhaseTracking(BaseSignalModel):
:param gamma: Gyromagnetic ratio in rad/ms/mT
"""
super().__init__(expand_repetitions=expand_repetitions, name="phase_tracking", **kwargs)
- self.wave_form = tf.Variable(wave_form, shape=(None, None, 4),
- dtype=tf.float32, name='wave_form')
- self.n_steps = tf.Variable(0, shape=(), dtype=tf.int64, name="n_steps")
- self.gamma = tf.constant(gamma, dtype=tf.float32)
- self.expansion_factor = tf.shape(self.wave_form)[0]
- self.update()
-
+ with tf.device(self.device):
+ self.wave_form = tf.Variable(wave_form, shape=(None, None, 4),
+ dtype=tf.float32, name='wave_form')
+ self.n_steps = tf.Variable(0, shape=(), dtype=tf.int64, name="n_steps")
+ self.gamma = tf.constant(gamma, dtype=tf.float32)
+ self.expansion_factor = tf.shape(self.wave_form)[0]
+ self.update()
+
+ # @tf.function(jit_compile=True)
def __call__(self, signal_tensor: tf.Tensor, trajectory: tf.Tensor, **kwargs) -> tf.Tensor:
""" Evaluates the phase integration for given trajectories
diff --git a/cmrsim/analytic/contrast/sequences.py b/cmrsim/analytic/contrast/sequences.py
index 3720180..3c0620b 100644
--- a/cmrsim/analytic/contrast/sequences.py
+++ b/cmrsim/analytic/contrast/sequences.py
@@ -50,12 +50,13 @@ class SpinEcho(BaseSignalModel):
f"of elements. Got {echo_time.shape} and {repetition_time.shape}")
super().__init__(expand_repetitions=expand_repetitions, name="spin_echo", **kwargs)
- self.TE = tf.Variable(echo_time, shape=(None, ), dtype=tf.float32, name='TE')
- self.TR = tf.Variable(repetition_time, shape=(None, ), dtype=tf.float32, name='TR')
- self.expansion_factor = self.TE.read_value().shape[0]
- self.update()
+ with tf.device(self.device):
+ self.TE = tf.Variable(echo_time, shape=(None, ), dtype=tf.float32, name='TE')
+ self.TR = tf.Variable(repetition_time, shape=(None, ), dtype=tf.float32, name='TR')
+ self.expansion_factor = self.TE.read_value().shape[0]
+ self.update()
- @tf.Module.with_name_scope
+ @tf.function(jit_compile=True)
def __call__(self, signal_tensor: tf.Tensor, T1: tf.Tensor, T2: tf.Tensor, **kwargs): # noqa
""" Evaluates spin echo operator
diff --git a/cmrsim/analytic/contrast/t2_star.py b/cmrsim/analytic/contrast/t2_star.py
index 72d5507..ebfa223 100644
--- a/cmrsim/analytic/contrast/t2_star.py
+++ b/cmrsim/analytic/contrast/t2_star.py
@@ -62,16 +62,17 @@ class StaticT2starDecay(BaseSignalModel):
f"(#repetitions, #k-space-points), but is: {sampling_times.shape}")
if len(sampling_times.shape) == 1:
sampling_times = tf.reshape(sampling_times, [1, -1])
-
- self.sampling_times = tf.Variable(tf.constant(sampling_times, tf.float32),
- shape=(None, None), dtype=tf.float32)
- self.expansion_factor = sampling_times.shape[0]
+ with tf.device(self.device):
+ self.sampling_times = tf.Variable(tf.constant(sampling_times, tf.float32),
+ shape=(None, None), dtype=tf.float32)
+ self.expansion_factor = sampling_times.shape[0]
def update(self):
super().update()
self.expansion_factor == tf.shape(self.sampling_times.read_value())[0]
# pylint: disable=signature-differs
+ @tf.function(jit_compile=True)
def __call__(self, signal_tensor: tf.Tensor, T2star: tf.Tensor, **kwargs):
""" Evaluates the T2* operator
--
GitLab
From e2e188b1d9ad0dd62caa4efd4ab029a498db0db7 Mon Sep 17 00:00:00 2001
From: Jonathan Weine <jweine@ethz.ch>
Date: Mon, 31 Jul 2023 16:21:24 +0200
Subject: [PATCH 2/3] Moved simulation k-segment loop over to base
implementation of encoding module
---
cmrsim/analytic/encoding/base.py | 64 +++++++++++++++++++-------
cmrsim/analytic/simulation.py | 77 ++++++++++++--------------------
2 files changed, 76 insertions(+), 65 deletions(-)
diff --git a/cmrsim/analytic/encoding/base.py b/cmrsim/analytic/encoding/base.py
index ea25c18..2eadff6 100644
--- a/cmrsim/analytic/encoding/base.py
+++ b/cmrsim/analytic/encoding/base.py
@@ -68,18 +68,19 @@ class BaseSampling(tf.Module):
if isinstance(absolute_noise_std, float):
absolute_noise_std = [absolute_noise_std, ]
- self.absolute_noise_std = tf.Variable(tf.constant(absolute_noise_std), dtype=tf.float32,
- trainable=False, shape=[None, ],
- name='absolute_noise_std')
- self.k_space_segments = tf.Variable(tf.constant(k_space_segments), dtype=tf.int32,
- trainable=False)
- self.k_space_vectors = tf.Variable(tf.zeros((1, 3), tf.float32), shape=[None, 3],
- dtype=tf.float32, trainable=False,
- name='k_space_vectors')
- self.sampling_times = tf.Variable(tf.zeros((1,), tf.float32), shape=[None, ],
- dtype=tf.float32, trainable=False, name='sampling_times')
- self.update()
- self.ori_matrix = orientation_matrix
+ with tf.device(self.device):
+ self.absolute_noise_std = tf.Variable(tf.constant(absolute_noise_std), dtype=tf.float32,
+ trainable=False, shape=[None, ],
+ name='absolute_noise_std')
+ self.k_space_segments = tf.Variable(tf.constant(k_space_segments), dtype=tf.int32,
+ trainable=False)
+ self.k_space_vectors = tf.Variable(tf.zeros((1, 3), tf.float32), shape=[None, 3],
+ dtype=tf.float32, trainable=False,
+ name='k_space_vectors')
+ self.sampling_times = tf.Variable(tf.zeros((1,), tf.float32), shape=[None, ],
+ dtype=tf.float32, trainable=False, name='sampling_times')
+ self.update()
+ self.ori_matrix = orientation_matrix
def update(self):
""" Assigns values to trajectory vectors. Should be used after modifying the parameters of
@@ -107,8 +108,39 @@ class BaseSampling(tf.Module):
self._segmented_k_vectors = tf.RaggedTensor.from_row_lengths(
self.k_space_vectors.read_value(), row_lengths)
- def __call__(self, transverse_magnetization: tf.Tensor, r_vectors: tf.Tensor,
- segment_index: Union[int, tf.Tensor] = 0, **kwargs):
+ @tf.function(jit_compile=False)
+ def __call__(self, m_transverse: tf.Tensor, r_vectors: tf.Tensor):
+ """ Fourier transforms the handed batch of object points / isochromates,
+
+ :param m_transverse:
+ :param r_vectors:
+ :return: tf.Tensor of shape (#repetitions, #k-space-samples)
+ """
+ with tf.device(self.device):
+ # Allocate k-space-Tensor
+ # s_of_k_segments = tf.TensorArray(dtype=tf.complex64, size=n_segments, dynamic_size=False)
+ samples_per_segment = tf.cast(self.number_of_samples / self.k_space_segments, tf.int32)
+ accumulator_shape = tf.stack([self.k_space_segments, tf.shape(m_transverse)[1], samples_per_segment], axis=0)
+ s_of_k_segments = tf.zeros(accumulator_shape, dtype=tf.complex64)
+ max_mtrans_index = tf.math.reduce_min(tf.stack([tf.shape(m_transverse)[-1], self.k_space_segments]))
+ max_rvectrans_index = tf.math.reduce_min(tf.stack([tf.shape(r_vectors)[-2], self.k_space_segments]))
+
+ # Loop over k-space segments, defined by k-space trajectory in self.encoding_module
+ for segment_index in tf.range(self.k_space_segments):
+ mseg_idx = tf.reduce_min(tf.stack([segment_index + 1, max_mtrans_index]))
+ rseg_idx = tf.reduce_min(tf.stack([segment_index + 1, max_rvectrans_index]))
+ m_segment = m_transverse[..., samples_per_segment * (mseg_idx - 1): samples_per_segment * mseg_idx]
+ r_segment = r_vectors[..., samples_per_segment * (rseg_idx - 1):samples_per_segment*rseg_idx, :]
+ k_space_segment = self.call_single_segment(m_segment, r_segment, segment_index=segment_index)
+ s_of_k_segments = tf.tensor_scatter_nd_update(s_of_k_segments, [[segment_index], ],
+ k_space_segment[tf.newaxis])
+
+ # Concat all segments and transpose axes to match needed format
+ return tf.reshape(tf.transpose(s_of_k_segments, [1, 0, 2]), [accumulator_shape[1], -1])
+
+ @tf.function(jit_compile=False)
+ def call_single_segment(self, transverse_magnetization: tf.Tensor, r_vectors: tf.Tensor,
+ segment_index: Union[int, tf.Tensor] = 0, **kwargs) -> tf.Tensor:
"""Calculates fourier phases for given object-representation at r-vectors. For multiple
different contrasts #repetitions is the representing axis.
@@ -122,10 +154,10 @@ class BaseSampling(tf.Module):
:param r_vectors: (#voxel, #repetitions, #k-space-samples, 3), axis #repetitions
and #k-space-samples can be 1 to broadcast for coordinate reuse.
:param segment_index: int
- :param kwargs: - noise_seed: if not None, sets seed for sampling the noise vector.
- :return: tf.Tensor
+ :return: tf.Tensor (..., k-space-points in segment)
"""
with tf.device(self.device):
+
n_repetitions = tf.shape(transverse_magnetization)[1]
fourier_factors = self._calculate_fourier_phases(r_vectors=r_vectors,
segment_index=segment_index)
diff --git a/cmrsim/analytic/simulation.py b/cmrsim/analytic/simulation.py
index fa1b08a..0ee93e1 100644
--- a/cmrsim/analytic/simulation.py
+++ b/cmrsim/analytic/simulation.py
@@ -15,6 +15,7 @@ if TYPE_CHECKING:
from cmrsim.utils.display import SimulationProgressBarII
+from time import perf_counter
class AnalyticSimulation(tf.Module):
""" This module provides the entry point to build and call the simulations defined within the
@@ -56,7 +57,7 @@ class AnalyticSimulation(tf.Module):
# Initialize progress bar
n_k_space_segments = self.encoding_module.k_space_segments.read_value()
self.progress_bar = SimulationProgressBarII(total_voxels=1, prefix='Run Simulation: ',
- total_segments=n_k_space_segments)
+ total_segments=n_k_space_segments)
@abstractmethod
def build(self, **kwargs) -> ('CompositeSignalModel', 'BaseSampling', 'BaseRecon'):
@@ -122,7 +123,7 @@ class AnalyticSimulation(tf.Module):
f'{map_names} does not contain all entries of'
f' {self.forward_model.required_quantities}')
- def _simulation_loop(self, dataset: tf.data.Dataset, k_space_shape: tf.Tensor,
+ def _simulation_loop(self, dataset: tf.data.Dataset, k_space_shape: tf.Tensor,
trajectory_module=None, trajectory_signatures: dict = None,
additional_kwargs: dict = None):
""" Consumes all object-configuration data (images) and simulates the MR images. The
@@ -137,64 +138,39 @@ class AnalyticSimulation(tf.Module):
(#repetitions, #samples). The returned images are only 3D if the
input data, encoding and recon module are 3D as well.
"""
+ if additional_kwargs is None:
+ additional_kwargs = {}
+
# Allocate tensor array to store the simulated images
s_of_k_temp = tf.zeros(k_space_shape, tf.complex64)
self.progress_bar.reset_voxel()
- s_of_k_temp *= (0. + 0.j)
# Loop over material points in dataset
for batch_idx, batch_dict in dataset.enumerate():
if trajectory_module is not None:
batch_dict = self._map_trajectories(batch_idx, batch_dict, trajectory_module,
trajectory_signatures, additional_kwargs)
- # print([f"{k}: {v.shape}" for k, v in batch_dict.items()])
- s_of_k_temp += self._segment_loop(**batch_dict)
+ m_transverse = self.forward_model(batch_dict["M0"], segment_index=0, **batch_dict)
+ s_of_k_temp += self.encoding_module(m_transverse, batch_dict["r_vectors"])
self.progress_bar.update(add_voxels=tf.shape(batch_dict['M0'])[0])
self.progress_bar.print()
-
self.progress_bar.print_final()
return s_of_k_temp
-
- @tf.function(jit_compile=False, reduce_retracing=True)
- def _segment_loop(self, **batch_dict):
- """ Fourier transforms the handed batch of object points / isochromates,
-
- :param batch_dict: dictionary of type {'property_name': tf.Tensor(...)}
- e.g. {'M0': tf.Tensor{}}
- :return: tf.Tensor of shape (1, #repetitions, #k-space-samples)
- """
- # Allocate k-space-Tensor
- n_segments = self.encoding_module.k_space_segments.read_value()
- s_of_k_segments = tf.TensorArray(dtype=tf.complex64, size=n_segments, dynamic_size=False)
- self.progress_bar.reset_segments()
-
- m_transverse = self.forward_model(batch_dict["M0"], segment_index=0, **batch_dict)
-
- samples_per_segment = tf.cast(self.encoding_module.number_of_samples /
- self.encoding_module.k_space_segments, tf.int32)
- max_mtrans_index = tf.math.reduce_min(tf.stack([tf.shape(m_transverse)[-1], n_segments]))
- max_rvectrans_index = tf.math.reduce_min(tf.stack([tf.shape(batch_dict['r_vectors'])[-2], n_segments]))
-
- # Loop over k-space segments, defined by k-space trajectory in self.encoding_module
- for segment_index in tf.range(n_segments):
- mseg_idx = tf.reduce_min(tf.stack([segment_index + 1, max_mtrans_index]))
- rseg_idx = tf.reduce_min(tf.stack([segment_index + 1, max_rvectrans_index]))
- m_segment = m_transverse[..., samples_per_segment * (mseg_idx - 1): samples_per_segment * mseg_idx]
- r_segment = batch_dict['r_vectors'][..., samples_per_segment * (rseg_idx - 1):samples_per_segment*rseg_idx, :]
- k_space_segment = self.encoding_module(m_segment, r_segment, segment_index=segment_index)
- k_space_segment = tf.transpose(k_space_segment, [1, 0])
- s_of_k_segments = s_of_k_segments.write(segment_index, k_space_segment)
-
- self.progress_bar.update(add_segment=1)
- if tf.math.floormod(segment_index, 5) == 0:
- self.progress_bar.print()
-
- # Concat all segments and transpose axes to match needed format
- return tf.transpose(s_of_k_segments.concat(), [1, 0])
@staticmethod
def _map_trajectories(idx: int, batch: dict, trajectory_module,
trajectory_signatures: dict, additional_kwargs: dict):
+ """
+
+ :warning: Only the last entry of the trajectory signatures dict
+
+ :param idx:
+ :param batch:
+ :param trajectory_module:
+ :param trajectory_signatures:
+ :param additional_kwargs:
+ :return:
+ """
init_r = batch.pop("initial_positions")
for k, t in trajectory_signatures.items():
new_shape = tf.concat([tf.shape(init_r)[0:1], tf.shape(t), [3, ]], axis=0)
@@ -202,17 +178,19 @@ class AnalyticSimulation(tf.Module):
# using the flattened time-samples
pos, add_lookups = [], []
for t_ in t:
- p, alup = trajectory_module(initial_positions=tf.reshape(init_r, (-1, 3)), timing=t_,
+ p, alup = trajectory_module(initial_positions=tf.reshape(init_r, (-1, 3)),
+ timing=tf.reshape(t_, (-1, )),
**additional_kwargs, batch_index=tf.cast(idx, tf.int32))
pos.append(p)
add_lookups.append(alup)
pos = tf.stack(pos, axis=1)
pos = tf.reshape(pos, new_shape)
- add_lookups = {k: tf.reshape(tf.stack([alup[k] for alup in add_lookups]),
- tf.concat([new_shape[:-1], tf.shape(add_lookups[0][k])]))
- for k in add_lookups[0].keys()}
-
batch.update({k: pos})
+
+ lookup_shapes = [tf.concat([new_shape[:-1], tf.shape(add_lookups[-1][k])[2:]], 0)
+ for k in add_lookups[-1].keys()]
+ add_lookups = {k: tf.reshape(tf.stack([alup[k] for alup in add_lookups]), lus)
+ for k, lus in zip(add_lookups[-1].keys(), lookup_shapes)}
batch.update(add_lookups)
return batch
@@ -294,7 +272,8 @@ class AnalyticSimulation(tf.Module):
print("Graph Tracing activated...")
tf.summary.trace_on(graph=True)
batch_dict, = [_ for _ in dataset(100).take(1)]
- _ = self._segment_loop(**batch_dict)
+ m_transverse = self.forward_model(batch_dict["M0"], segment_index=0, **batch_dict)
+ _ = self.encoding_module(m_transverse, batch_dict["r_vectors"])
with writer.as_default(): # pylint: disable not-context-manager
tf.summary.trace_export(name='graph_def', step=0)
--
GitLab
From f4bbc398141ee667a5ba109a7fc2fb22b1c8183e Mon Sep 17 00:00:00 2001
From: Jonathan Weine <jweine@ethz.ch>
Date: Mon, 31 Jul 2023 17:11:00 +0200
Subject: [PATCH 3/3] Removed unused and undocumented modules
---
cmrsim/analytic/encoding/base.py | 87 +-------------------------------
tests/test_encoding_base.py | 28 +---------
2 files changed, 3 insertions(+), 112 deletions(-)
diff --git a/cmrsim/analytic/encoding/base.py b/cmrsim/analytic/encoding/base.py
index 2eadff6..c4ac13b 100644
--- a/cmrsim/analytic/encoding/base.py
+++ b/cmrsim/analytic/encoding/base.py
@@ -1,5 +1,5 @@
""" This module contains the base implementation of the encoding mechanism of the simulation"""
-__all__ = ["BaseSampling", "InternalAccumulator", "BlochOperatorInput"]
+__all__ = ["BaseSampling",]
from typing import Union, Iterable, Tuple
import abc
@@ -278,87 +278,4 @@ class BaseSampling(tf.Module):
self.ori_matrix = tf.Variable(trafo_mat.astype(np.float32), shape=(3, 4),
dtype=tf.float32)
else:
- self.ori_matrix.assign(trafo_mat.astype(np.float32))
-
-
-# pylint: disable=abstract-method
-class InternalAccumulator(tf.Module):
- """Wraps a BaseSampling object to enable internal k-space accumulation"""
- #: Encoding module that is wrapped
- wrapped_module: BaseSampling
- _kspace_acc: Tuple[tf.Variable]
-
- def __init__(self, encoding_module: BaseSampling, n_repetitions: int):
- """
- :param encoding_module: Instance of an encoding module
- :param n_repetitions: number of repetitions to be expected for
- given transversal magnetization
- """
- self.wrapped_module = encoding_module
- super().__init__(name=encoding_module.name)
- self.update(n_repetitions)
-
- def update(self, n_repetitions: int):
- self.wrapped_module.update()
- # pylint: disable=protected-access
- samples_per_segment = [tf.shape(_)[0] for _ in
- self.wrapped_module._segmented_sampling_times]
-
- with tf.device("CPU:0"):
- # pylint: disable=consider-using-generator
- self._kspace_acc = tuple([tf.Variable(tf.zeros((n_repetitions, s),
- dtype=tf.complex64),
- dtype=tf.complex64, shape=(None, s))
- for s in samples_per_segment])
-
- def __getattr__(self, attr):
- """ This is only called if attr is not in self.__dict__"""
- if attr != "wrapped_module":
- return getattr(self.wrapped_module, attr)
-
- def __call__(self, transverse_magnetization: tf.Tensor, r_vectors: tf.Tensor,
- segment_index: Union[int, tf.Tensor] = 0, **kwargs):
- samples_batch = self.wrapped_module(transverse_magnetization, r_vectors,
- segment_index, **kwargs)
- with tf.device("CPU:0"):
- self._kspace_acc[segment_index].assign_add(samples_batch)
- return transverse_magnetization
-
- def reset(self):
- """ Sets all k-space-line accumulators to zero """
- for i, accumulator in enumerate(self._kspace_acc):
- self._kspace_acc[i].assign(tf.zeros_like(accumulator))
-
- def get_kspace(self):
- """ Returns k-space data in shape (1, -1, #samples)"""
- return tf.concat(self._kspace_acc, -1)
-
-
-# pylint: disable=abstract-method
-class BlochOperatorInput(tf.Module):
- """Wraps a BaseSampling object to deal with the difference in input shape for bloch operators
- and solution operators """
-
- def __init__(self, encoding_module: Union[BaseSampling, InternalAccumulator]):
- """
- :param encoding_module: Instance of an encoding module
- """
- super().__init__(name=encoding_module.name)
- self.wrapped_module = encoding_module
-
- def __getattr__(self, attr):
- """ This is only called if attr is not in self.__dict__"""
- if attr != "wrapped_module":
- return getattr(self.wrapped_module, attr)
-
- def __call__(self, magnetization: tf.Tensor, trajectories: tf.Tensor,
- segment_index: Union[int, tf.Tensor] = 0, **kwargs):
- """
-
- :param magnetization: (#batch, 3)
- :param trajectories: (#batch, #k-space-points, 3)
- """
- m_transverse = tf.reshape(magnetization[:, 0], [-1, 1, 1])
- rvec = tf.reshape(trajectories, [magnetization.shape[0], 1, -1, 3])
- return self.wrapped_module(transverse_magnetization=m_transverse, r_vectors=rvec,
- segment_index=segment_index, **kwargs)
+ self.ori_matrix.assign(trafo_mat.astype(np.float32))
\ No newline at end of file
diff --git a/tests/test_encoding_base.py b/tests/test_encoding_base.py
index e5fc449..6e65000 100644
--- a/tests/test_encoding_base.py
+++ b/tests/test_encoding_base.py
@@ -5,7 +5,7 @@ import tensorflow as tf
import numpy as np
from pint import Quantity
-from cmrsim.analytic.encoding import BaseSampling, InternalAccumulator, BlochOperatorInput, EPI
+from cmrsim.analytic.encoding import BaseSampling, EPI
class TestInputShapes(unittest.TestCase):
@@ -50,32 +50,6 @@ class TestInputShapes(unittest.TestCase):
k = self.module(m_transverse, r_vectors)
self.assertTrue(all(i == j for i, j in zip(target_shape, tuple(k.shape))))
- def test_internal_accumulator(self):
- actual_number_of_contrasts = 10
- n_voxels = 1000
- shape_list = itertools.product([1, actual_number_of_contrasts], [1, self.n_sampling_points])
- module = InternalAccumulator(self.module, n_repetitions=actual_number_of_contrasts)
-
- for reps, timings in shape_list:
- with self.subTest():
- target_shape = (actual_number_of_contrasts, self.n_sampling_points)
- m_transverse = tf.ones((n_voxels, actual_number_of_contrasts, timings),
- dtype=tf.complex64)
- r_vectors = tf.random.normal((n_voxels, reps, timings, 3))
- module(m_transverse, r_vectors, 0)
- k = module.get_kspace()
- self.assertTrue(all(i == j for i, j in zip(target_shape, tuple(k.shape))))
-
- def test_bloch_input_handler(self):
- module = BlochOperatorInput(self.module)
- m_transverse = tf.ones((100, 3), dtype=tf.complex64)
- with self.subTest():
- r_vectors = tf.random.normal((100, 1, 3))
- k = module(m_transverse, r_vectors)
- with self.subTest():
- r_vectors = tf.random.normal((100, module._segmented_sampling_times[0].shape[0], 3))
- k = module(m_transverse, r_vectors)
-
class TestValues(unittest.TestCase):
def test_dc_input(self):
--
GitLab