def _Build(self) -> BatchData:
"""Construct and return a batch, resetting mutable state."""
gt = self.builder.Build()
# Expand node labels to 1-hot.
indices = np.arange(len(self.node_labels))
node_labels_1hot = np.zeros((len(self.node_labels), 2), dtype=np.int32)
node_labels_1hot[indices, self.node_labels] = 1
batch = BatchData(
graph_count=gt.graph_size,
model_data=GgnnBatchData(
graph_tuple=gt,
vocab_ids=np.array(self.vocab_ids, dtype=np.int32),
selector_ids=np.array(self.selector_ids, dtype=np.int32),
node_labels=node_labels_1hot,
),
)
# Reset mutable state.
self.vocab_ids = []
self.selector_ids = []
self.node_labels = []
self.node_size = 0
return batch
def _Build(self) -> BatchData:
# A batch may contain fewer graphs than the required batch_size.
# If so, pad with empty "graphs". These padding graphs will be discarded
# once processed.
if len(self.graph_node_sizes) < self.batch_size:
pad_count = self.batch_size - len(self.graph_node_sizes)
self.vocab_ids += [np.array([self._vocab_id_pad], dtype=np.int32)
] * pad_count
self.selector_vectors += [self._selector_vector_pad] * pad_count
self.targets += [self._node_label_pad] * pad_count
batch = BatchData(
graph_count=len(self.graph_node_sizes),
model_data=LstmBatchData(
graph_node_sizes=np.array(self.graph_node_sizes,
dtype=np.int32),
encoded_sequences=tf.compat.v1.keras.preprocessing.sequence.
pad_sequences(
self.vocab_ids,
maxlen=self.padded_sequence_length,
dtype="int32",
padding="pre",
truncating="post",
value=self._vocab_id_pad,
),
selector_vectors=tf.compat.v1.keras.preprocessing.sequence.
pad_sequences(
self.selector_vectors,
maxlen=self.padded_sequence_length,
dtype="float32",
padding="pre",
truncating="post",
value=np.zeros(2, dtype=np.float32),
),
node_labels=tf.compat.v1.keras.preprocessing.sequence.
pad_sequences(
self.targets,
maxlen=self.padded_sequence_length,
dtype="float32",
padding="pre",
truncating="post",
value=np.zeros(self.node_y_dimensionality,
dtype=np.float32),
),
# We don't pad or truncate targets.
targets=self.targets,
),
)
# Reset mutable state.
self.graph_node_sizes = []
self.vocab_ids = []
self.selector_vectors = []
self.targets = []
return batch
def test_RollingResults_iteration_count(weight: float):
"""Test aggreation of model iteration count and convergence."""
rolling_results = RollingResults()
data = BatchData(graph_count=1, model_data=None)
results = BatchResults.Create(
targets=np.array([[0, 1, 2]]),
predictions=np.array([[0, 1, 2]]),
iteration_count=1,
model_converged=True,
)
for _ in range(10):
rolling_results.Update(data, results, weight=weight)
assert rolling_results.iteration_count == 1
assert rolling_results.model_converged == 1