PythonNinja.md

Python Ninja

Set of tricks to become a true Master of Python. Directly related with Data Science and ML tools such as tensorflow, pandas and common interfaces between them.

1. Dictionary where keys can be accessed as attributes:

    class AttrDict(dict):
      __getattr__ = dict.__getitem__
      __setattr__ = dict.__setitem__
   

   One example could be:

      config = AttrDict()
      config.layer_sizes = [100, 100]
      config.output_size = 10
      config.learning_rate = 1e-3
   

   The difference with an empty class with attributes? We have access to all the nice dict functions and properties.

2. Decorator to share variables across functions and methods in TF: link to GIST. (Code also found in this repo under code/share_variables.py

3. Return all TF variables defined in a function:

    def define_graph(config):
      tf.reset_default_graph()
      inputs, targets = define_data_pipeline()
      prediction = my_network(inputs, config)
      loss = tf.losses.mean_squared_error(targets, prediction)
      optimizer = config.optimizer()
      optimize = optimizer.minimize(loss)
      return AttrDict(locals())  # The magic line.
   

   Thanks to this, we can define a graph with the following way:

    graph = define_graph(config)
    with tf.Session() as sess:
      sess.run(tf.global_variables_initializer())
      for _ in range(config.num_epochs):
        sess.run(graph.optimize)
      loss = sess.run(graph.loss)  # No name collision anymore.
      print(loss)