MCPcopy Create free account
hub / github.com/google-research/language / main

Function main

language/orqa/preprocessing/preprocess_wiki_extractor.py:73–102  ·  view source on GitHub ↗
(_)

Source from the content-addressed store, hash-verified

71
72
73def main(_):
74 pool = multiprocessing.Pool(FLAGS.num_threads)
75 tf.logging.info("Using hub module %s", FLAGS.bert_hub_module_path)
76 tokenizer = bert_utils.get_tokenizer(FLAGS.bert_hub_module_path)
77 preprocessor = wiki_preprocessor.Preprocessor(get_sentence_splitter(),
78 FLAGS.max_block_length,
79 tokenizer)
80 mapper = functools.partial(create_block_info, preprocessor=preprocessor)
81 block_count = 0
82 input_paths = tf.io.gfile.glob(FLAGS.input_pattern)
83 random.shuffle(input_paths)
84 tf.logging.info("Processing %d input files.", len(input_paths))
85
86 tf.io.gfile.makedirs(FLAGS.output_dir)
87 blocks_path = os.path.join(FLAGS.output_dir, "blocks.tfr")
88 examples_path = os.path.join(FLAGS.output_dir, "examples.tfr")
89 titles_path = os.path.join(FLAGS.output_dir, "titles.tfr")
90
91 with tf.python_io.TFRecordWriter(blocks_path) as blocks_writer:
92 with tf.python_io.TFRecordWriter(examples_path) as examples_writer:
93 with tf.python_io.TFRecordWriter(titles_path) as titles_writer:
94 for block_info in pool.imap_unordered(mapper, input_paths):
95 for title, block, examples in block_info:
96 blocks_writer.write(block.encode("utf-8"))
97 examples_writer.write(examples)
98 titles_writer.write(title.encode("utf-8"))
99 block_count += 1
100 if block_count % 10000 == 0:
101 tf.logging.info("Wrote %d blocks.", block_count)
102 tf.logging.info("Wrote %d blocks in total.", block_count)
103
104if __name__ == "__main__":
105 app.run(main)

Callers

nothing calls this directly

Calls 3

get_sentence_splitterFunction · 0.85
get_tokenizerMethod · 0.45
encodeMethod · 0.45

Tested by

no test coverage detected

Used in the wild real call sites across dependent graphs

searching dependent graphs…