TensorFlow / Keras多线程模型拟合

时间:2017-02-19 01:35:18

标签: multithreading concurrency tensorflow keras python-multithreading

我尝试使用多个线程(以及keras后端)训练具有不同参数值的多个tensorflow模型。我已经看到了在多个线程中使用相同模型的一些示例,但在这种特殊情况下,我遇到了有关冲突图等的各种错误。这里是我喜欢的一个简单示例能够做到:

from concurrent.futures import ThreadPoolExecutor
import numpy as np
import tensorflow as tf
from keras import backend as K
from keras.layers import Dense
from keras.models import Sequential


sess = tf.Session()


def example_model(size):
    model = Sequential()
    model.add(Dense(size, input_shape=(5,)))
    model.add(Dense(1))
    model.compile(optimizer='sgd', loss='mse')
    return model


if __name__ == '__main__':
    K.set_session(sess)
    X = np.random.random((10, 5))
    y = np.random.random((10, 1))
    models = [example_model(i) for i in range(5, 10)]

    e = ThreadPoolExecutor(4)
    res_list = [e.submit(model.fit, X, y) for model in models]

    for res in res_list:
        print(res.result())

结果错误为ValueError: Tensor("Variable:0", shape=(5, 5), dtype=float32_ref) must be from the same graph as Tensor("Variable_2/read:0", shape=(), dtype=float32).。我也尝试在线程中初始化模型,这会导致类似的失败。

关于最佳方式的任何想法?我完全不依赖于这个确切的结构,但我更喜欢能够使用多个线程而不是进程,因此所有模型都在相同的GPU内存分配中进行训练。

1 个答案:

答案 0 :(得分:17)

Tensorflow图不是线程安全的(参见https://www.tensorflow.org/api_docs/python/tf/Graph),当您创建新的Tensorflow会话时,它默认使用默认图。

您可以通过在并行化函数中创建一个包含新图形的新会话并在那里构建keras模型来解决这个问题。

以下是一些代码,它们可以并行地为每个可用的gpu创建和拟合模型:

import concurrent.futures
import numpy as np

import keras.backend as K
from keras.layers import Dense
from keras.models import Sequential

import tensorflow as tf
from tensorflow.python.client import device_lib

def get_available_gpus():
    local_device_protos = device_lib.list_local_devices()
    return [x.name for x in local_device_protos if x.device_type == 'GPU']

xdata = np.random.randn(100, 8)
ytrue = np.random.randint(0, 2, 100)

def fit(gpu):
    with tf.Session(graph=tf.Graph()) as sess:
        K.set_session(sess)
        with tf.device(gpu):
            model = Sequential()
            model.add(Dense(12, input_dim=8, activation='relu'))
            model.add(Dense(8, activation='relu'))
            model.add(Dense(1, activation='sigmoid'))

            model.compile(loss='binary_crossentropy', optimizer='adam')
            model.fit(xdata, ytrue, verbose=0)

            return model.evaluate(xdata, ytrue, verbose=0)

gpus = get_available_gpus()
with concurrent.futures.ThreadPoolExecutor(len(gpus)) as executor:
    results = [x for x in executor.map(fit, gpus)]
print('results: ', results)