def run_stats(graph_path: RichPath, output_path: RichPath):
number_graphs, number_annotations, number_variables = 0, 0, 0
annotation_table = Counter()
data_generator = chain(
*(g.read_as_jsonl()
for g in graph_path.iterate_filtered_files_in_dir('*.jsonl.gz')))
for data in data_generator:
number_graphs += 1 if len(data['supernodes']) > 0 else 0
number_variables += len(data['supernodes'])
number_annotations += sum(
1 for supernode in data['supernodes'].values()
if supernode['annotation'] not in {None, 'None', 'Nothing', 'Any'})
annotation_table.update((supernode['annotation']
for supernode in data['supernodes'].values()
if supernode['annotation'] not in
{None, 'None', 'Nothing', 'Any'}))
with open(output_path.to_local_path().path, "a") as f:
f.write("Statistics for file: " + graph_path.to_local_path().path +
"\n")
f.write("Number of graphs: %d\n" % (number_graphs))
f.write("Number of variables: %d\n" % (number_variables))
f.write("Number of annotations: %d\n" % (number_annotations))
f.write("Number of different annotations: %d\n" %
(len(list(annotation_table))))
f.write("\nFrequency distribution of annotations type:\n\n")
for annotation, value in annotation_table.most_common():
f.write("%s\t%d\n" % (annotation, value))
def restore_model(cls, path: RichPath, device=None) -> 'BaseComponent':
"""Restore model to a given device."""
model_path = path.to_local_path().path
with gzip.open(model_path, 'rb') as f:
model = torch.load(f, map_location=device) # type: BaseComponent
if device is not None:
model.to(device)
return model
