-
Notifications
You must be signed in to change notification settings - Fork 488
/
tensorflow2_keras_mnist.py
94 lines (78 loc) · 3.84 KB
/
tensorflow2_keras_mnist.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
89
90
91
92
93
94
# Copyright 2020 Uber Technologies, Inc. All Rights Reserved.
# Copyright 2019 Uber Technologies, Inc. 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.
# ==============================================================================
from __future__ import absolute_import, division, print_function
import tensorflow as tf
import byteps.tensorflow.keras as bps
# tf.compat.v1.disable_eager_execution()
# byteps: initialize byteps.
bps.init()
# byteps: pin GPU to be used to process local rank (one GPU per process)
gpus = tf.config.experimental.list_physical_devices('GPU')
for gpu in gpus:
tf.config.experimental.set_memory_growth(gpu, True)
if gpus:
tf.config.experimental.set_visible_devices(gpus[bps.local_rank()], 'GPU')
(mnist_images, mnist_labels), _ = \
tf.keras.datasets.mnist.load_data(path='mnist-%d.npz' % bps.rank())
dataset = tf.data.Dataset.from_tensor_slices(
(tf.cast(mnist_images[..., tf.newaxis] / 255.0, tf.float32),
tf.cast(mnist_labels, tf.int64))
)
dataset = dataset.repeat().shuffle(10000).batch(128)
mnist_model = tf.keras.Sequential([
tf.keras.layers.Conv2D(32, [3, 3], activation='relu'),
tf.keras.layers.Conv2D(64, [3, 3], activation='relu'),
tf.keras.layers.MaxPooling2D(pool_size=(2, 2)),
tf.keras.layers.Dropout(0.25),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(128, activation='relu'),
tf.keras.layers.Dropout(0.5),
tf.keras.layers.Dense(10, activation='softmax')
])
# byteps: adjust learning rate based on number of GPUs.
scaled_lr = 0.001 * bps.size()
opt = tf.optimizers.Adam(scaled_lr)
# byteps: add byteps DistributedOptimizer.
opt = bps.DistributedOptimizer(opt)
# byteps: Specify `experimental_run_tf_function=False` to ensure TensorFlow
# uses bps.DistributedOptimizer() to compute gradients.
mnist_model.compile(loss=tf.losses.SparseCategoricalCrossentropy(),
optimizer=opt,
metrics=['accuracy'],
experimental_run_tf_function=False)
callbacks = [
# byteps: broadcast initial variable states from rank 0 to all other processes.
# This is necessary to ensure consistent initialization of all workers when
# training is started with random weights or restored from a checkpoint.
bps.callbacks.BroadcastGlobalVariablesCallback(0, device="GPU:0"),
# byteps: average metrics among workers at the end of every epoch.
#
# Note: This callback must be in the list before the ReduceLROnPlateau,
# TensorBoard or other metrics-based callbacks.
bps.callbacks.MetricAverageCallback(device="GPU:0"),
# byteps: using `lr = 1.0 * bps.size()` from the very beginning leads to worse final
# accuracy. Scale the learning rate `lr = 1.0` ---> `lr = 1.0 * bps.size()` during
# the first three epochs. See https://arxiv.org/abs/1706.02677 for details.
bps.callbacks.LearningRateWarmupCallback(warmup_epochs=3, initial_lr=scaled_lr, verbose=1),
]
# byteps: save checkpoints only on worker 0 to prevent other workers from corrupting them.
if bps.rank() == 0:
callbacks.append(tf.keras.callbacks.ModelCheckpoint('./checkpoint-{epoch}.h5'))
# byteps: write logs on worker 0.
verbose = 1 if bps.rank() == 0 else 0
# Train the model.
# byteps: adjust number of steps based on number of GPUs.
mnist_model.fit(dataset, steps_per_epoch=500 // bps.size(), callbacks=callbacks, epochs=24, verbose=verbose)