2d. Distributed training and monitoring

In this notebook, we refactor to call train_and_evaluate instead of hand-coding our ML pipeline. This allows us to carry out evaluation as part of our training loop instead of as a separate step. It also adds in failure-handling that is necessary for distributed training capabilities.

We also use TensorBoard to monitor the training.

In [1]:
from google.cloud import bigquery
import tensorflow as tf
import numpy as np
import shutil
print(tf.__version__)
1.15.2

Input

Read data created in Lab1a, but this time make it more general, so that we are reading in batches. Instead of using Pandas, we will use add a filename queue to the TensorFlow graph.

In [2]:
CSV_COLUMNS = ['fare_amount', 'pickuplon','pickuplat','dropofflon','dropofflat','passengers', 'key']
LABEL_COLUMN = 'fare_amount'
DEFAULTS = [[0.0], [-74.0], [40.0], [-74.0], [40.7], [1.0], ['nokey']]

def read_dataset(filename, mode, batch_size = 512):
      def decode_csv(value_column):
          columns = tf.decode_csv(value_column, record_defaults = DEFAULTS)
          features = dict(zip(CSV_COLUMNS, columns))
          label = features.pop(LABEL_COLUMN)
          # No need to features.pop('key') since it is not specified in the INPUT_COLUMNS.
          # The key passes through the graph unused.
          return features, label

      # Create list of file names that match "glob" pattern (i.e. data_file_*.csv)
      filenames_dataset = tf.data.Dataset.list_files(filename)
      # Read lines from text files
      textlines_dataset = filenames_dataset.flat_map(tf.data.TextLineDataset)
      # Parse text lines as comma-separated values (CSV)
      dataset = textlines_dataset.map(decode_csv)

      # Note:
      # use tf.data.Dataset.flat_map to apply one to many transformations (here: filename -> text lines)
      # use tf.data.Dataset.map      to apply one to one  transformations (here: text line -> feature list)

      if mode == tf.estimator.ModeKeys.TRAIN:
          num_epochs = None # indefinitely
          dataset = dataset.shuffle(buffer_size = 10 * batch_size)
      else:
          num_epochs = 1 # end-of-input after this

      dataset = dataset.repeat(num_epochs).batch(batch_size)
      
      return dataset

Create features out of input data

For now, pass these through. (same as previous lab)

In [3]:
INPUT_COLUMNS = [
    tf.feature_column.numeric_column('pickuplon'),
    tf.feature_column.numeric_column('pickuplat'),
    tf.feature_column.numeric_column('dropofflat'),
    tf.feature_column.numeric_column('dropofflon'),
    tf.feature_column.numeric_column('passengers'),
]

def add_more_features(feats):
    # Nothing to add (yet!)
    return feats

feature_cols = add_more_features(INPUT_COLUMNS)

Serving input function

In [4]:
# Defines the expected shape of the JSON feed that the model
# will receive once deployed behind a REST API in production.
def serving_input_fn():
    json_feature_placeholders = {
        'pickuplon' : tf.placeholder(tf.float32, [None]),
        'pickuplat' : tf.placeholder(tf.float32, [None]),
        'dropofflat' : tf.placeholder(tf.float32, [None]),
        'dropofflon' : tf.placeholder(tf.float32, [None]),
        'passengers' : tf.placeholder(tf.float32, [None]),
    }
    # You can transforma data here from the input format to the format expected by your model.
    features = json_feature_placeholders # no transformation needed
    return tf.estimator.export.ServingInputReceiver(features, json_feature_placeholders)

tf.estimator.train_and_evaluate

In [5]:
def train_and_evaluate(output_dir, num_train_steps):
    estimator = tf.estimator.LinearRegressor(
                       model_dir = output_dir,
                       feature_columns = feature_cols)
    
    train_spec=tf.estimator.TrainSpec(
                       input_fn = lambda: read_dataset('./taxi-train.csv', mode = tf.estimator.ModeKeys.TRAIN),
                       max_steps = num_train_steps)

    exporter = tf.estimator.LatestExporter('exporter', serving_input_fn)

    eval_spec=tf.estimator.EvalSpec(
                       input_fn = lambda: read_dataset('./taxi-valid.csv', mode = tf.estimator.ModeKeys.EVAL),
                       steps = None,
                       start_delay_secs = 1, # start evaluating after N seconds
                       throttle_secs = 10,  # evaluate every N seconds
                       exporters = exporter)
    
    tf.estimator.train_and_evaluate(estimator, train_spec, eval_spec)

Monitor training with TensorBoard

To activate TensorBoard within the JupyterLab UI navigate to "File" - "New Launcher". Then double-click the 'Tensorboard' icon on the bottom row.

TensorBoard 1 will appear in the new tab. Navigate through the three tabs to see the active TensorBoard. The 'Graphs' and 'Projector' tabs offer very interesting information including the ability to replay the tests.

You may close the TensorBoard tab when you are finished exploring.

In [6]:
OUTDIR = './taxi_trained'

Run training

In [ ]:
# Run training    
shutil.rmtree(OUTDIR, ignore_errors = True) # start fresh each time
tf.summary.FileWriterCache.clear() # ensure filewriter cache is clear for TensorBoard events file
train_and_evaluate(OUTDIR, num_train_steps = 500)
INFO:tensorflow:Using default config.
INFO:tensorflow:Using config: {'_model_dir': './taxi_trained', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': allow_soft_placement: true
graph_options {
  rewrite_options {
    meta_optimizer_iterations: ONE
  }
}
, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_session_creation_timeout_secs': 7200, '_service': None, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x7f6c0b5e64d0>, '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}
INFO:tensorflow:Not using Distribute Coordinator.
INFO:tensorflow:Running training and evaluation locally (non-distributed).
INFO:tensorflow:Start train and evaluate loop. The evaluate will happen after every checkpoint. Checkpoint frequency is determined based on RunConfig arguments: save_checkpoints_steps None or save_checkpoints_secs 600.
WARNING:tensorflow:From /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/training/training_util.py:236: Variable.initialized_value (from tensorflow.python.ops.variables) is deprecated and will be removed in a future version.
Instructions for updating:
Use Variable.read_value. Variables in 2.X are initialized automatically both in eager and graph (inside tf.defun) contexts.
WARNING:tensorflow:From /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/autograph/converters/directives.py:119: The name tf.decode_csv is deprecated. Please use tf.io.decode_csv instead.

INFO:tensorflow:Calling model_fn.
WARNING:tensorflow:From /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/feature_column/feature_column_v2.py:305: Layer.add_variable (from tensorflow.python.keras.engine.base_layer) is deprecated and will be removed in a future version.
Instructions for updating:
Please use `layer.add_weight` method instead.
WARNING:tensorflow:From /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/ops/resource_variable_ops.py:1630: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.
Instructions for updating:
If using Keras pass *_constraint arguments to layers.
WARNING:tensorflow:From /opt/conda/lib/python3.7/site-packages/tensorflow_estimator/python/estimator/canned/linear.py:308: to_float (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.cast` instead.
INFO:tensorflow:Done calling model_fn.
INFO:tensorflow:Create CheckpointSaverHook.
WARNING:tensorflow:From /opt/conda/lib/python3.7/site-packages/tensorflow_core/python/ops/array_ops.py:1475: where (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.where in 2.0, which has the same broadcast rule as np.where
INFO:tensorflow:Graph was finalized.
INFO:tensorflow:Running local_init_op.
INFO:tensorflow:Done running local_init_op.
INFO:tensorflow:Saving checkpoints for 0 into ./taxi_trained/model.ckpt.
INFO:tensorflow:loss = 98185.664, step = 1

Challenge Exercise

Modify your solution to the challenge exercise in c_dataset.ipynb appropriately.

Copyright 2017 Google Inc. 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