python - 如何保存/恢复tensorflow的tensor_forest模型？

Question

我使用 tensorflow 来运行随机森林模型。代码:

import tensorflow as tf
from tensorflow.contrib.tensor_forest.python import tensor_forest
from tensorflow.python.ops import resources
from tensorflow.examples.tutorials.mnist import input_data

num_steps = 50000 # Total steps to train
batch_size = 1024 # The number of samples per batch
num_classes = 10 # The 10 digits
num_features = 784 # Each image is 28x28 pixels
num_trees = 10
max_nodes = 1000

X = tf.placeholder(tf.float32, shape=[None, num_features])
Y = tf.placeholder(tf.int32, shape=[None])
hparams = tensor_forest.ForestHParams(num_classes=num_classes,
                                     num_features=num_features,
                                     num_trees=num_trees,
                                     max_nodes=max_nodes).fill()
forest_graph = tensor_forest.RandomForestGraphs(params=hparams)
train_op = forest_graph.training_graph(X, Y)
loss_op = forest_graph.training_loss(X,Y)
infer_op = forest_graph.inference_graph(X)
correct_prediction = tf.equal(tf.arg_max(infer_op, 1), tf.cast(Y, tf.int64))
accuracy_op = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

init_vars =  tf.group(tf.global_variables_initializer(), resources.initialize_resources(resources.shared_resources()))
sess = tf.Session()
sess.run(init_vars)
test_x, test_y = mnist.test.images, mnist.test.labels
for i in range(1, num_steps + 1):
    batch_x, batch_y = mnist.train.next_batch(batch_size=batch_size)
    _, l = sess.run([train_op, loss_op], feed_dict={X:batch_x, Y: batch_y})
    if i % 100 == 0 or i == 1:
        acc = sess.run(accuracy_op, feed_dict={X:batch_x, Y: batch_y})
        print('step %i, loss: %f, acc: %f' % (i, l, acc))
    if i % 100 == 0:
        print("Test Accuracy:", sess.run(accuracy_op, feed_dict={X: test_x, Y: test_y}))
print("Test Accuracy:", sess.run(accuracy_op, feed_dict={X: test_x, Y: test_y}))

问题:如何保存模型并恢复模型进行预测？这是tf随机森林的最新版本，我使用tf 1.2，它可以工作。我发现有人使用 TensorForestEstimator，但它不适用于 tf 1.2，tf更新这么频繁!

Answer 1

保存模型很容易，但恢复它会杀了我。无论我做什么，总是出现“FertileStatsResourceHandleOp”错误，最后，我在恢复之前添加两行代码，它起作用了。

hparams = tensor_forest.ForestHParams(num_classes=num_classes,
                                     num_features=num_features,
                                     num_trees=num_trees,
                                     max_nodes=max_nodes).fill()
forest_graph = tensor_forest.RandomForestGraphs(params=hparams)

完整代码如下:

X = tf.placeholder(tf.float32, shape=[None, num_features],name="input_x")
Y = tf.placeholder(tf.int32, shape=[None], name="input_y")
hparams = tensor_forest.ForestHParams(num_classes=num_classes,
                                     num_features=num_features,
                                     num_trees=num_trees,
                                     max_nodes=max_nodes).fill()
forest_graph = tensor_forest.RandomForestGraphs(params=hparams)
train_op = forest_graph.training_graph(X, Y)
loss_op = forest_graph.training_loss(X,Y)

correct_prediction = tf.argmax(infer_op, 1, name="predictions")
accuracy_op = tf.reduce_mean(tf.cast(tf.equal(correct_prediction,tf.cast(Y, tf.int64)), tf.float32),name="accuracy")

init_vars =  tf.group(tf.global_variables_initializer(), resources.initialize_resources(resources.shared_resources()))
sess = tf.Session()
sess.run(init_vars)
test_x, test_y = mnist.test.images, mnist.test.labels
saver = tf.train.Saver(save_relative_paths=True, max_to_keep=10)
checkpoint_prefix = 'checkpoints/model'
for i in range(1, num_steps + 1):
    batch_x, batch_y = mnist.train.next_batch(batch_size=batch_size)
    _, l = sess.run([train_op, loss_op], feed_dict={X:batch_x, Y: batch_y})
    if i % 10 == 0 or i == 1:
        acc = sess.run(accuracy_op, feed_dict={X:batch_x, Y: batch_y})
        print('step %i, loss: %f, acc: %f' % (i, l, acc))
    if i % 10 == 0:
        print("Test Accuracy:", sess.run(accuracy_op, feed_dict={X: test_x, Y: test_y}))
        path = saver.save(sess, checkpoint_prefix, global_step=i)
        print("last Saved model checkpoint to {} at step {}".format(path, i))
print("Test Accuracy:", sess.run(accuracy_op, feed_dict={X: test_x, Y: test_y}))

恢复模型:

hparams = tensor_forest.ForestHParams(num_classes=num_classes,
                                     num_features=num_features,
                                     num_trees=num_trees,
                                     max_nodes=max_nodes).fill()
forest_graph = tensor_forest.RandomForestGraphs(params=hparams)
checkpoint_file = tf.train.latest_checkpoint('checkpoints')
graph = tf.Graph()
with graph.as_default():
    session_conf = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
    sess = tf.Session(config=session_conf)
    with sess.as_default():
        saver = tf.train.import_meta_graph("{}.meta".format(checkpoint_file), clear_devices=True)
        saver.restore(sess, checkpoint_file)
        input_x = graph.get_operation_by_name("input_x").outputs[0]
        input_y = graph.get_operation_by_name("input_y").outputs[0]
        predictions = graph.get_operation_by_name("predictions").outputs[0]
        accuracy = graph.get_operation_by_name("accuracy").outputs[0]
        acc = sess.run(accuracy, {input_x: test_x, input_y:test_y })
        predictions = sess.run(predictions, {input_x: test_x })
        print(predictions)