def on_batch_end(self, data: Data) -> None:
if self.write_graph and self.system.network.epoch_models.symmetric_difference(
self.painted_graphs):
self.writer.write_epoch_models(mode=self.system.mode)
self.painted_graphs = self.system.network.epoch_models
if self.system.mode != 'train':
return
if self.histogram_freq.freq and self.histogram_freq.is_step and \
self.system.global_step % self.histogram_freq.freq == 0:
self.writer.write_weights(mode=self.system.mode,
models=self.system.network.models,
step=self.system.global_step,
visualize=self.paint_weights)
if self.update_freq.freq and self.update_freq.is_step and self.system.global_step % self.update_freq.freq == 0:
self.writer.write_scalars(mode=self.system.mode,
step=self.system.global_step,
scalars=filter(lambda x: is_number(x[1]),
data.items()))
self.writer.write_images(mode=self.system.mode,
step=self.system.global_step,
images=filter(
lambda x: x[1] is not None,
map(lambda y: (y, data.get(y)),
self.write_images)))
self.writer.write_embeddings(
mode=self.system.mode,
step=self.system.global_step,
embeddings=filter(
lambda x: x[1] is not None,
map(
lambda t:
(t[0], data.get(t[0]), data.get(t[1]), data.get(t[2])),
self.write_embeddings)))
def on_epoch_end(self, data: Data) -> None:
if self.system.mode == 'train' and self.histogram_freq.freq and not self.histogram_freq.is_step and \
self.system.epoch_idx % self.histogram_freq.freq == 0:
self.writer.write_weights(mode=self.system.mode,
models=self.system.network.models,
step=self.system.global_step,
visualize=self.paint_weights)
if self.update_freq.freq and (self.update_freq.is_step
or self.system.epoch_idx %
self.update_freq.freq == 0):
self.writer.write_scalars(mode=self.system.mode,
step=self.system.global_step,
scalars=filter(lambda x: is_number(x[1]),
data.items()))
self.writer.write_images(mode=self.system.mode,
step=self.system.global_step,
images=filter(
lambda x: x[1] is not None,
map(lambda y: (y, data.get(y)),
self.write_images)))
self.writer.write_embeddings(
mode=self.system.mode,
step=self.system.global_step,
embeddings=filter(
lambda x: x[1] is not None,
map(
lambda t:
(t[0], data.get(t[0]), data.get(t[1]), data.get(t[2])),
self.write_embeddings)))
def load_and_interpret(model_path, input_paths, baseline=-1, input_type='float32', dictionary_path=None,
strip_alpha=False, smooth_factor=7, save=False, save_dir=None):
""" A helper class to load input and invoke the interpretation api
Args:
model_path: The path the model file (str)
input_paths: The paths to model input files [(str),...]
baseline: Either a number corresponding to the baseline for integration, or a path to a baseline file
input_type: The data type of the model inputs, ex 'float32'
dictionary_path: The path to a dictionary file encoding a 'class_idx'->'class_name' mapping
strip_alpha: Whether to collapse alpha channels when loading an input (bool)
smooth_factor: How many iterations of the smoothing algorithm to run (int)
save: Whether to save (True) or display (False) the resulting image
save_dir: Where to save the image if save=True
"""
model_dir = os.path.dirname(model_path)
if save_dir is None:
save_dir = model_dir
network = keras.models.load_model(model_path)
dic = load_dict(dictionary_path)
inputs = [load_image(input_paths[i], strip_alpha=strip_alpha) for i in range(len(input_paths))]
max_shapes = np.maximum.reduce([inp.shape for inp in inputs], axis=0)
tf_image = tf.stack([tf.image.resize_with_crop_or_pad(tf.convert_to_tensor(im, dtype=input_type), max_shapes[0],
max_shapes[1]) for im in inputs], axis=0)
if is_number(baseline):
baseline_gen = tf.constant_initializer(float(baseline))
baseline_image = baseline_gen(shape=tf_image.shape, dtype=input_type)
else:
baseline_image = load_image(baseline)
baseline_image = tf.convert_to_tensor(baseline_image, dtype=input_type)
interpret_model(network, tf_image, baseline_input=baseline_image, decode_dictionary=dic, smooth=smooth_factor,
save=save, save_path=save_dir)
def load_and_saliency(model_path,
input_paths,
baseline=-1,
dictionary_path=None,
strip_alpha=False,
smooth_factor=7,
save=False,
save_dir=None):
"""A helper class to load input and invoke the saliency api
Args:
model_path: The path the model file (str)
input_paths: The paths to model input files [(str),...] or to a folder of inputs [(str)]
baseline: Either a number corresponding to the baseline for integration, or a path to a baseline file
dictionary_path: The path to a dictionary file encoding a 'class_idx'->'class_name' mapping
strip_alpha: Whether to collapse alpha channels when loading an input (bool)
smooth_factor: How many iterations of the smoothing algorithm to run (int)
save: Whether to save (True) or display (False) the resulting image
save_dir: Where to save the image if save=True
"""
model_dir = os.path.dirname(model_path)
if save_dir is None:
save_dir = model_dir
if not save:
save_dir = None
network = keras.models.load_model(model_path, compile=False)
input_type = network.input.dtype
input_shape = network.input.shape
n_channels = 0 if len(input_shape) == 3 else input_shape[3]
dic = load_dict(dictionary_path)
if len(input_paths) == 1 and os.path.isdir(input_paths[0]):
loader = PathLoader(input_paths[0])
input_paths = [path[0] for path in loader.path_pairs]
inputs = [
load_image(input_paths[i],
strip_alpha=strip_alpha,
channels=n_channels) for i in range(len(input_paths))
]
max_shapes = np.maximum.reduce([inp.shape for inp in inputs], axis=0)
tf_image = tf.stack([
tf.image.resize_with_crop_or_pad(
tf.convert_to_tensor(im, dtype=input_type), max_shapes[0],
max_shapes[1]) for im in inputs
],
axis=0)
if is_number(baseline):
baseline_gen = tf.constant_initializer(float(baseline))
baseline_image = baseline_gen(shape=tf_image.shape, dtype=input_type)
else:
baseline_image = load_image(baseline)
baseline_image = tf.convert_to_tensor(baseline_image, dtype=input_type)
visualize_saliency(network,
tf_image,
baseline_input=baseline_image,
decode_dictionary=dic,
smooth=smooth_factor,
save_path=save_dir)
def _infer_keys(self, state):
monitored_keys = []
for key, val in state.items():
if isinstance(val, str) or is_number(val):
monitored_keys.append(key)
elif hasattr(val, "numpy") and len(val.numpy().shape) == 1:
monitored_keys.append(key)
self.keys = sorted(monitored_keys)
def on_epoch_end(self, data: Data) -> None:
if self.system.mode == 'train' and self.histogram_freq.freq and not self.histogram_freq.is_step and \
self.system.epoch_idx % self.histogram_freq.freq == 0:
self.writer.write_weights(mode=self.system.mode,
models=self.system.network.models,
step=self.system.global_step,
visualize=self.paint_weights)
# Write out any embeddings which were aggregated over batches
for name, val_list in self.collected_embeddings.items():
embeddings = None if any(
x[0] is None
for x in val_list) else concat([x[0] for x in val_list])
labels = None if any(
x[1] is None
for x in val_list) else concat([x[1] for x in val_list])
imgs = None if any(
x[2] is None
for x in val_list) else concat([x[2] for x in val_list])
self.writer.write_embeddings(mode=self.system.mode,
step=self.system.global_step,
embeddings=[(name, embeddings, labels,
imgs)])
self.collected_embeddings.clear()
# Get any embeddings which were generated externally on epoch end
if self.embedding_freq.freq and (self.embedding_freq.is_step
or self.system.epoch_idx %
self.embedding_freq.freq == 0):
self.writer.write_embeddings(
mode=self.system.mode,
step=self.system.global_step,
embeddings=filter(
lambda x: x[1] is not None,
map(
lambda t:
(t[0], data.get(t[0]), data.get(t[1]), data.get(t[2])),
self.write_embeddings)))
if self.update_freq.freq and (self.update_freq.is_step
or self.system.epoch_idx %
self.update_freq.freq == 0):
self.writer.write_scalars(mode=self.system.mode,
step=self.system.global_step,
scalars=filter(lambda x: is_number(x[1]),
data.items()))
self.writer.write_images(mode=self.system.mode,
step=self.system.global_step,
images=filter(
lambda x: x[1] is not None,
map(lambda y: (y, data.get(y)),
self.write_images)))
def on_epoch_end(self, state):
with self.summary_writers[state['mode']].as_default():
for key in state.keys() - self.ignore_keys:
val = state[key]
if is_number(val):
tf.summary.scalar("epoch_" + key, val, step=state['epoch'])
for key in self.write_images - {True, False}:
data = state.get(key)
if data is not None:
tf.summary.image(key, data, step=state['epoch'])
if state['mode'] == 'train' and self.histogram_freq and state[
'epoch'] % self.histogram_freq == 0:
self._log_weights(epoch=state['epoch'])
if state['mode'] == 'train' and self.embeddings_freq and state[
'epoch'] % self.embeddings_freq == 0:
self._log_embeddings(state)
def on_batch_end(self, data: Data) -> None:
if self.write_graph and self.system.network.epoch_models.symmetric_difference(
self.painted_graphs):
self.writer.write_epoch_models(mode=self.system.mode,
epoch=self.system.epoch_idx)
self.painted_graphs = self.system.network.epoch_models
# Collect embeddings if present in batch but viewing per epoch. Don't aggregate during training though
if self.system.mode != 'train' and self.embedding_freq.freq and not self.embedding_freq.is_step and self.system.epoch_idx % self.embedding_freq.freq == 0:
for elem in self.write_embeddings:
name, lbl, img = elem
if name in data:
self.collected_embeddings[name].append(
(data.get(name), data.get(lbl), data.get(img)))
# Handle embeddings if viewing per step
if self.embedding_freq.freq and self.embedding_freq.is_step and self.system.global_step % self.embedding_freq.freq == 0:
self.writer.write_embeddings(
mode=self.system.mode,
step=self.system.global_step,
embeddings=filter(
lambda x: x[1] is not None,
map(
lambda t:
(t[0], data.get(t[0]), data.get(t[1]), data.get(t[2])),
self.write_embeddings)))
if self.system.mode != 'train':
return
if self.histogram_freq.freq and self.histogram_freq.is_step and \
self.system.global_step % self.histogram_freq.freq == 0:
self.writer.write_weights(mode=self.system.mode,
models=self.system.network.models,
step=self.system.global_step,
visualize=self.paint_weights)
if self.update_freq.freq and self.update_freq.is_step and self.system.global_step % self.update_freq.freq == 0:
self.writer.write_scalars(mode=self.system.mode,
step=self.system.global_step,
scalars=filter(lambda x: is_number(x[1]),
data.items()))
self.writer.write_images(mode=self.system.mode,
step=self.system.global_step,
images=filter(
lambda x: x[1] is not None,
map(lambda y: (y, data.get(y)),
self.write_images)))
def on_batch_end(self, state):
if state['mode'] != 'train':
return
if self.is_tracing:
self._log_trace(state['train_step'])
elif state['train_step'] == self.profile_batch - 1:
self._enable_trace()
if self.update_freq == 'epoch' or state[
'train_step'] % self.update_freq != 0:
return
with self.summary_writers[state['mode']].as_default():
for key in state.keys() - self.ignore_keys:
val = state[key]
if is_number(val):
tf.summary.scalar("batch_" + key,
val,
step=state['train_step'])
for key in self.write_images - {True, False}:
data = state.get(key)
if data is not None:
tf.summary.image(key, data, step=state['train_step'])
