lab_getting_started.py

#

Lab ๐Ÿงช

This library lets you organize TensorFlow machine learning experiments.

It is based on a bunch of utility functions and classes I wrote while trying some machine learning algorithms. I recently made it to a separate repo because I’ve been reusing them on different projects, and it was easier to keep track of them as a single project.

โš ๏ธ This project is a personal project and only a few weeks old. It still doesn’t have a proper documentation and I’m making improvements frequently. So, if you wish to use it make sure you understand the code. I have tried to keep it documented and simple.

What does it do?

  • Keeps checkpoints and TensorBoard summaries and logs organized
  • Helps keep track of experiments with reference to git commits
  • Produce pretty console outputs
  • Maintains and logs histograms and moving averages
  • Monitor time taken for different sections of code
  • Estimate time remaining for experiments to run
  • Add headers to python source files with experiment results for easy reference
  • A simple TensorBoard invoker
  • Tools for custom graphs from TensorBoard summaries
  • Help make code more readable

Why I made it?

I started coding existing reinforcement learning algorithms to play Atari games for fun. It was not easy to keep track of things when I started trying variations, fixing bugs etc. This library helps organize experiments. It organizes the folders of the checkpoints, logs and TensorBoard summaries by each experiment. It also keeps track of the git commits when each experiment was run, so if some other change in code, affected the results of a experiment you can easily track what caused it.

I also wrote a logger to display pretty results on screen and to make it easy to write TensorBoard summaries. It also keeps track of training times which makes it easy to spot what’s taking up most resources. Here’s the output of this sample program (lab_getting_started.py):

Updates

  • November 16, 2018

    • Initial public release
  • December 21, 2018

    • TensorBoard invoker
    • Tool set for custom visualizations of TensorBoard summaries
    • Automatically adds headers to python files with experiment results

This ๐Ÿ‘‡ is the header that was added automatically when the experiment ran.

2018-12-21 09:46:41
Sample lab experiment
 [dirty](dirty.html): ๐Ÿ“š tb
start_step: 0

-------------------------------------
| global_step |   reward |     loss |
-------------------------------------
|           9 |     1.50 |    13.50 |
|          19 |     4.83 |    23.50 |
|          29 |     8.17 |    33.50 |
|          39 |    11.50 |    43.50 |
-------------------------------------
#

Here’s how to use it

#

Time module is used to sleep so that you can see how this sample works when you run it.

import lab.clear_warnings
import time

import tensorflow as tf

from lab.experiment import Experiment
#

You should keep the project level lab details defined in a python file at the top of the project. Here’s the example lab_globals.py used for this sample.

from lab_globals import lab
#

Create the sample experiment

EXPERIMENT = Experiment(lab=lab,
                        name="Sample",
                        python_file=__file__,
                        comment="Sample lab experiment",
                        check_repo_dirty=False)
#

Get a reference to logger

logger = EXPERIMENT.logger
#

This is sample monitored section. I use monitored sections to keep track of what’s going on from the console output. It is also useful to organize the code into sections, when separating them into functions is difficult

with logger.monitor("Create model") as m:
#

Indicate that this section failed. You don’t have to set this if it is successful.

    m.is_successful = False
#

Sleep for a minute.

    time.sleep(1)
#

Print sample info

logger.info(one=1,
            two=2,
            string="string")
#

Set logger indicators

#

Reward is a queue; this is useful when you want to track the moving average of something.

logger.add_indicator("reward", queue_limit=10)
#

By default everything is a set of values and will create a TensorBoard histogram We specify that fps is a scalar. If you store multiple values for this it will output the mean.

logger.add_indicator("fps", is_histogram=False, is_progress=False)
#

This will produce a histogram

logger.add_indicator("loss")
#

Create a TensorFlow session

with tf.Session() as session:
#

Start the experiment from scratch, without loading from a saved checkpoint. This will clear all the old checkpoints and summaries for this experiment. If you start with the continued non-zero global_step the experiment will load from the last saved checkpoint.

    EXPERIMENT.start(0, session)
#

Create monitored iterator

    monitor = logger.iterator(range(1000))
#

This is the main training loop of this project.

    for global_step in monitor:
#

Handle Keyboard Interrupts

        try:
            with logger.delayed_keyboard_interrupt():
#

Print the step

                logger.print_global_step(global_step)
#

A sample monitored section inside iterator

                with monitor.section("sample"):
                    time.sleep(0.5)
#

An unmonitored section is used only to organize code. It produces no output

                with monitor.unmonitored("logging"):
#

Store a dictionary

                    logger.store(
                        reward=global_step / 3.0,
                        fps=12
                    )
#

Store a collection of values

                    for i in range(global_step, global_step + 10):
                        logger.store(loss=i)
#

Another monitored section

                with monitor.section("process_samples"):
                    time.sleep(0.5)
#

A third monitored section to make it real

                with monitor.section("train"):
#

Let it run for multiple iterations. We’ll track the progress of that too

                    iterations = 100
                    progress = logger.progress(iterations)
                    for i in range(100):
                        time.sleep(0.01)
#

Progress is tracked manually unlike in the top level iterator The progress updates do not have to be sequential.

                        progress.update(i)
#

Clears the progress when complete

                    progress.clear()
#

Store progress in the trials file and in the python code as a comment

                if (global_step + 1) % 10 == 0:
                    progress_dict = logger.get_progress_dict(global_step=global_step)
#

Set is_add to False if you want to update the last progress record instead of making a new entry

                    EXPERIMENT.save_progress(progress_dict, is_add=True)
#

Log stored values. This will output to the console and write TensorBoard summaries.

                logger.write(global_step=global_step, new_line=False)
#

Show iterator progress. This will display how much time is remaining.

                monitor.progress()
#

We will overwrite the same console line, and produce a new line after ten such overwrites. This helps keep the console output concise.

                logger.clear_line(reset=(global_step + 1) % 10 != 0)
        except KeyboardInterrupt:
            logger.log(f"Stopping the training at {global_step} and saving checkpoints")
            break
#

TensorBoard invoker

This lets you start TensorBoard without having to type in all the log paths. For instance, so that you can start it with

python tools/tb.py -e ppo ppo_transformed_bellman

instead of

tensorboard --log_dir=ppo:project_path/logs/ppo,ppo_transformed_bellman:project_path/logs/ppo_transformed_bellman

To get a list of all available experiments

python tools/tb.py -e ppo ppo_transformed_bellman

Custom analysis of TensorBoard summaries

TensorBoard is nice, but sometimes you need custom charts to debug algorithms. Following is an example of a custom chart:

And sometime TensorBoard is not even doing a good job; for instance lets say you have a histogram, with 90% of data points between 1 and 2 whilst there are a few outliers at 1000 - you won’t be able to see the distribution between 1 and 2 because the graph is scaled to 1000.

I think TensorBoard will develop itself to handle these. And the main reason behind these tooling I’ve written is for custom charts, and because it’s not that hard to do it.

Here’s a link to a sample Jupyter Notebook with custom charts.

github repo

@vpj on Twitter