forked from tensorflow/models
-
Notifications
You must be signed in to change notification settings - Fork 2
/
export_albert_tfhub.py
88 lines (74 loc) · 3.38 KB
/
export_albert_tfhub.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""A script to export the ALBERT core model as a TF-Hub SavedModel."""
from __future__ import absolute_import
from __future__ import division
# from __future__ import google_type_annotations
from __future__ import print_function
from absl import app
from absl import flags
import tensorflow as tf
from typing import Text
from official.nlp.albert import configs
from official.nlp.bert import bert_models
FLAGS = flags.FLAGS
flags.DEFINE_string("albert_config_file", None,
"Albert configuration file to define core albert layers.")
flags.DEFINE_string("model_checkpoint_path", None,
"File path to TF model checkpoint.")
flags.DEFINE_string("export_path", None, "TF-Hub SavedModel destination path.")
flags.DEFINE_string(
"sp_model_file", None,
"The sentence piece model file that the ALBERT model was trained on.")
def create_albert_model(
albert_config: configs.AlbertConfig) -> tf.keras.Model:
"""Creates an ALBERT keras core model from ALBERT configuration.
Args:
albert_config: An `AlbertConfig` to create the core model.
Returns:
A keras model.
"""
# Adds input layers just as placeholders.
input_word_ids = tf.keras.layers.Input(
shape=(None,), dtype=tf.int32, name="input_word_ids")
input_mask = tf.keras.layers.Input(
shape=(None,), dtype=tf.int32, name="input_mask")
input_type_ids = tf.keras.layers.Input(
shape=(None,), dtype=tf.int32, name="input_type_ids")
transformer_encoder = bert_models.get_transformer_encoder(
albert_config, sequence_length=None)
sequence_output, pooled_output = transformer_encoder(
[input_word_ids, input_mask, input_type_ids])
# To keep consistent with legacy hub modules, the outputs are
# "pooled_output" and "sequence_output".
return tf.keras.Model(
inputs=[input_word_ids, input_mask, input_type_ids],
outputs=[pooled_output, sequence_output]), transformer_encoder
def export_albert_tfhub(albert_config: configs.AlbertConfig,
model_checkpoint_path: Text, hub_destination: Text,
sp_model_file: Text):
"""Restores a tf.keras.Model and saves for TF-Hub."""
core_model, encoder = create_albert_model(albert_config)
checkpoint = tf.train.Checkpoint(model=encoder)
checkpoint.restore(model_checkpoint_path).assert_consumed()
core_model.sp_model_file = tf.saved_model.Asset(sp_model_file)
core_model.save(hub_destination, include_optimizer=False, save_format="tf")
def main(_):
albert_config = configs.AlbertConfig.from_json_file(
FLAGS.albert_config_file)
export_albert_tfhub(albert_config, FLAGS.model_checkpoint_path,
FLAGS.export_path, FLAGS.sp_model_file)
if __name__ == "__main__":
app.run(main)