In [1]:

    

import os
os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"]="2"


    

In [ ]:

    

from tensorflow.contrib.learn.python.learn import metric_spec
from tensorflow.contrib.learn.python.learn.estimators import estimator
from tensorflow.contrib.tensor_forest.client import eval_metrics
from tensorflow.contrib.tensor_forest.client import random_forest
from tensorflow.contrib.tensor_forest.python import tensor_forest

from tensorflow.examples.tutorials.mnist import input_data
#from tensorflow.python.platform import app

FLAGS = None

batch_size = 128
model_dir = '/tmp/mnist/'
num_trees = 10
max_nodes = 10
use_training_loss = True

def build_estimator(model_dir):
    """Build an estimator."""
    params = tensor_forest.ForestHParams(
        num_classes=10, num_features=784,
        num_trees=num_trees, max_nodes=max_nodes)
    
    graph_builder_class = tensor_forest.RandomForestGraphs
    
    if use_training_loss:
        graph_builder_class = tensor_forest.TrainingLossForest
    
    # Use the SKCompat wrapper, which gives us a convenient way to split
    # in-memory data like MNIST into batches.
    return estimator.SKCompat(random_forest.TensorForestEstimator(
        params, graph_builder_class=graph_builder_class,
        model_dir=model_dir))


"""Train and evaluate the model."""
est = build_estimator(model_dir)

mnist = input_data.read_data_sets('/tmp/mnist/', one_hot=False)

est.fit(x=mnist.train.images, y=mnist.train.labels, batch_size=batch_size)


    

In [8]:

    

metric_name = 'accuracy'
metric = {metric_name:
            metric_spec.MetricSpec(
                eval_metrics.get_metric(metric_name),
                prediction_key=eval_metrics.get_prediction_key(metric_name))}

results = est.score(x=mnist.test.images, y=mnist.test.labels,
                      batch_size=batch_size,
                      metrics=metric)
for key in sorted(results):
    print('%s: %s' % (key, results[key]))


    

    




---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-8-8ca7c1347771> in <module>()
      7 results = est.score(x=mnist.test.images, y=mnist.test.labels,
      8                       batch_size=batch_size,
----> 9                       metrics=metric)
     10 for key in sorted(results):
     11     print('%s: %s' % (key, results[key]))

/home/jorge/anaconda3/envs/keras2_py36/lib/python3.6/site-packages/tensorflow/contrib/learn/python/learn/estimators/estimator.py in score(self, x, y, batch_size, steps, metrics)
   1361       raise ValueError('Metrics argument should be None or dict. '
   1362                        'Got %s.' % metrics)
-> 1363     eval_results, global_step = self._estimator._evaluate_model(
   1364         input_fn=input_fn,
   1365         feed_fn=feed_fn,

AttributeError: 'TensorForestEstimator' object has no attribute '_evaluate_model'

In [ ]: