pyoneer

Tensor utilities, reinforcement learning, and more! Designed to make research easier with low-level abstractions for common operations.

Usage

For the top-level utilities, import like so:

import pyoneer as pynr
pynr.math.rescale(...)

For the larger sub-modules, such as reinforcement learning, we recommend:

import pyoneer.rl as pyrl
loss_fn = pyrl.losses.PolicyGradient(...)

In general, the Pyoneer API tries to adhere to the TensorFlow 2.0 API.

Examples

• Proximal Policy Optimization with Pyoneer and TF 2.0

API

Activations (pynr.activations)

• pynr.activations.swish

Debugging (pynr.debugging)

• pynr.debugging.Stopwatch

Distributions (pynr.distributions)

• pynr.distributions.MultiCategorical

Initializers (pynr.initializers)

• pynr.initializers.SoftplusInverse

Layers (pynr.layers)

• pynr.layers.Swish
• pynr.layers.OneHotEncoder
• pynr.layers.AngleEncoder
• pynr.layers.Nest
• pynr.layers.ConcreteDropout

Tensor Manipulation (pynr.manip)

• pynr.manip.flatten
• pynr.manip.batched_index
• pynr.manip.pad_or_truncate
• pynr.manip.shift

Math (pynr.math)

• pynr.math.to_radians
• pynr.math.to_degrees
• pynr.math.to_cartesian
• pynr.math.to_polar
• pynr.math.RADIANS_TO_DEGREES
• pynr.math.DEGREES_TO_RADIANS
• pynr.math.isclose
• pynr.math.safe_divide
• pynr.math.rescale
• pynr.math.normalize
• pynr.math.denormalize
• pynr.math.sigmoid_inverse

Metrics (pynr.metrics)

• pynr.metrics.mape
• pynr.metrics.smape
• pynr.metrics.MAPE
• pynr.metrics.SMAPE

Moments (pynr.moments)

• pynr.moments.range_moments
• pynr.moments.StaticMoments
• pynr.moments.StreamingMoments
• pynr.moments.ExponentialMovingMoments

Learning Rate Schedules (pynr.schedules)

• pynr.schedules.CyclicSchedule

Regularizers (pynr.regularizers)

• pynr.regularizers.DropoutL2

Reinforcement Learning (pynr.rl)

Utilities for reinforcement learning.

Losses (pynr.rl.losses)

• pynr.rl.losses.policy_gradient
• pynr.rl.losses.policy_entropy
• pynr.rl.losses.clipped_policy_gradient
• pynr.rl.losses.PolicyGradient
• pynr.rl.losses.PolicyEntropy
• pynr.rl.losses.ClippedPolicyGradient

Targets (pynr.rl.targets)

• pynr.rl.targets.DiscountedReturns
• pynr.rl.targets.GeneralizedAdvantages

Strategies (pynr.rl.strategies)

• pynr.rl.strategies.EpsilonGreedy
• pynr.rl.strategies.Mode
• pynr.rl.strategies.Sample

Wrappers (pynr.rl.wrappers)

• pynr.rl.wrappers.ObservationCoordinates
• pynr.rl.wrappers.ObservationNormalization
• pynr.rl.wrappers.Batch
• pynr.rl.wrappers.Process

Installation

There are a few options for installation:

1. (Recommended) Install with pipenv:

    pipenv install fomoro-pyoneer
   

2. Install locally for development with pipenv:

    git clone https://github.com/fomorians/pyoneer.git
    cd pyoneer
    pipenv install
    pipenv shell
   

Testing

There are a few options for testing:

1. Run all tests:

    python -m unittest discover -bfp '*_test.py'
   

2. Run specific tests:

    python -m pyoneer.math.logical_ops_test
   

Contributing

File an issue following the ISSUE_TEMPLATE. If the issue discussion warrants implementation, then submit a pull request from a branch describing the feature. This will eventually get merged into master after a few rounds of code review.