tensorflow-estimator

All the official google tutorials use the one shot iterator for all the estimator api implementation, i couldnt find any documentation on how to use tf.data's initializable iterator and reinitializable interator instead of one shot iterator. Can someone kindly show me how to switch between train_data and test_data using tf.data's initializable iterator and reinitializable interator. We need to run a session to use feed dict and switch the dataset in the initializable iterator, its a low level api and its confusing how to use it part of estimator api architecture PS : I did find that google

We recently switched to Keras for TF 2.0, but when we compared it to the DNNClassifier Estimator on 2.0, we experienced around 4x slower speeds with Keras. But I cannot for the life of me figure out why this is happening. The rest of the code for both are identical, using an input_fn() that returns the same tf.data.Dataset, and using identical feature_columns. Been struggling with this problem for days now. Any help would be greatly greatly appreciated. Thank you Estimator code: estimator = tf.estimator.DNNClassifier( feature_columns = feature_columns, hidden_units = [64,64], activation_fn =

I have the following input_fn. def input_fn(filenames, batch_size): # Create a dataset containing the text lines. dataset = tf.data.TextLineDataset(filenames).skip(1) # Parse each line. dataset = dataset.map(_parse_line) # Shuffle, repeat, and batch the examples. dataset = dataset.shuffle(10000).repeat().batch(batch_size) # Return the dataset. return dataset It works great if filenames=['file1.csv'] or filenames=['file2.csv'] . It gives me an error if filenames=['file1.csv', 'file2.csv'] . In Tensorflow documentation it says filenames is a tf.string tensor containing one or more filenames. How

I am using tf.estimator.Estimator for developing my model, I wrote a model_fn and trained 50,000 iterations, now I want to make a small change in my model_fn , for example add a new layer. I don't want to start training from scratch, I want to restore all the old variables from the 50,000 checkpoint, and continue training from this point. When I try to do so I get a NotFoundError How can this be done with tf.estimator.Estimator ? TL;DR The easiest way to load variables from a previous checkpoint is to use the function tf.train.init_from_checkpoint() . Just one call to this function inside the

I am working with Tensorflow version 1.4, and I want to debug my train() function. In this link https://www.tensorflow.org/programmers_guide/debugger#debugging_tf-learn_estimators_and_experiments there is a way to do it for tf.contrib.learn Estimators , but I can not find a way to adapt it to the (new in version 1.4) tf.estimator . This is what I have tried: from tensorflow.python import debug as tf_debug # Create an estimator my_estimator = tf.estimator.Estimator(model_fn=model_fn, params=model_params, model_dir='/tb_dir', config=config_estimator) # Create a LocalCLIDebugHook and use it as a

I want to use feedable iterator design in tensorflow Dataset API, so I can switch to validation data after some training steps. But if I switched to validation data, it will end the whole session. The following code demonstrate what I want to do: import tensorflow as tf graph = tf.Graph() with graph.as_default(): training_ds = tf.data.Dataset.range(32).batch(4) validation_ds = tf.data.Dataset.range(8).batch(4) handle = tf.placeholder(tf.string, shape=[]) iterator = tf.data.Iterator.from_string_handle( handle, training_ds.output_types, training_ds.output_shapes) next_element = iterator.get_next

I'm trying to do transfer learning of an Inception-resnet v2 model pretrained on imagenet, using my own dataset and classes. My original codebase was a modification of a tf.slim sample which I can't find anymore and now I'm trying to rewrite the same code using the tf.estimator.* framework. I am running, however, into the problem of loading only some of the weights from the pretrained checkpoint, initializing the remaining layers with their default initializers. Researching the problem, I found this GitHub issue and this question , both mentioning the need to use tf.train.init_from_checkpoint

I am using tf.estimator.Estimator to train a model: def model_fn(features, labels, mode, params, config): input_image = features["input_image"] eval_metric_ops = {} predictions = {} # Create model with tf.name_scope('Model'): W = tf.Variable(tf.zeros([784, 10]), name="W") b = tf.Variable(tf.zeros([10]), name="b") logits = tf.nn.softmax(tf.matmul(input_image, W, name="MATMUL") + b, name="logits") loss = None train_op = None if mode != tf.estimator.ModeKeys.PREDICT: loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=labels, logits=logits)) train_op = tf.contrib.layers.optimize