Fast Discounted Cumulative Sums in PyTorch

Overview

TODO: update this README!

Fast Discounted Cumulative Sums in PyTorch

PyPiVersion PythonVersion PyPiDownloads License License: CC BY 4.0 Twitter Follow

This repository implements an efficient parallel algorithm for the computation of discounted cumulative sums and a Python package with differentiable bindings to PyTorch. The discounted cumsum operation is frequently seen in data science domains concerned with time series, including Reinforcement Learning (RL).

The traditional sequential algorithm performs the computation of the output elements in a loop. For an input of size N, it requires O(N) operations and takes O(N) time steps to complete.

The proposed parallel algorithm requires a total of O(N log N) operations, but takes only O(log N) time steps, which is a considerable trade-off in many applications involving large inputs.

Features of the parallel algorithm:

  • Speed logarithmic in the input size
  • Better numerical precision than sequential algorithms

Features of the package:

  • CPU: sequential algorithm in C++
  • GPU: parallel algorithm in CUDA
  • Gradients computation wrt input
  • Both left and right directions of summation supported
  • PyTorch bindings

Usage

Installation

pip install torch-discounted-cumsum

API

  • discounted_cumsum_right: Computes discounted cumulative sums to the right of each position (a standard setting in RL)
  • discounted_cumsum_left: Computes discounted cumulative sums to the left of each position

Example

import torch
from torch_discounted_cumsum import discounted_cumsum_right

N = 8
gamma = 0.99
x = torch.ones(1, N).cuda()
y = discounted_cumsum_right(x, gamma)

print(y)

Output:

tensor([[7.7255, 6.7935, 5.8520, 4.9010, 3.9404, 2.9701, 1.9900, 1.0000]],
       device='cuda:0')

Up to K elements

import torch
from torch_discounted_cumsum import discounted_cumsum_right

N = 8
K = 2
gamma = 0.99
x = torch.ones(1, N).cuda()
y_N = discounted_cumsum_right(x, gamma)
y_K = y_N - (gamma ** K) * torch.cat((y_N[:, K:], torch.zeros(1, K).cuda()), dim=1)

print(y_K)

Output:

tensor([[1.9900, 1.9900, 1.9900, 1.9900, 1.9900, 1.9900, 1.9900, 1.0000]],
       device='cuda:0')

Parallel Algorithm

For the sake of simplicity, the algorithm is explained for N=16. The processing is performed in-place in the input vector in log2 N stages. Each stage updates N / 2 positions in parallel (that is, in a single time step, provided unrestricted parallelism). A stage is characterized by the size of the group of sequential elements being updated, which is computed as 2 ^ (stage - 1). The group stride is always twice larger than the group size. The elements updated during the stage are highlighted with the respective stage color in the figure below. Here input elements are denoted with their position id in hex, and the elements tagged with two symbols indicate the range over which the discounted partial sum is computed upon stage completion.

Each element update includes an in-place addition of a discounted element, which follows the last updated element in the group. The discount factor is computed as gamma raised to the power of the distance between the updated and the discounted elements. In the figure below, this operation is denoted with tilted arrows with a greek gamma tag. After the last stage completes, the output is written in place of the input.

In the CUDA implementation, N / 2 CUDA threads are allocated during each stage to update the respective elements. The strict separation of updates into stages via separate kernel invocations guarantees stage-level synchronization and global consistency of updates.

The gradients wrt input can be obtained from the gradients wrt output by simply taking cumsum operation with the reversed direction of summation.

Numerical Precision

The parallel algorithm produces a more numerically-stable output than the sequential algorithm using the same scalar data type.

The comparison is performed between 3 runs with identical inputs (code). The first run casts inputs to double precision and obtains the output reference using the sequential algorithm. Next, we run both sequential and parallel algorithms with the same inputs cast to single precision and compare the results to the reference. The comparison is performed using the L_inf norm, which is just the maximum of per-element discrepancies.

With 10000-element non-zero-centered input (such as all elements are 1.0), the errors of the algorithms are 2.8e-4 (sequential) and 9.9e-5 (parallel). With zero-centered inputs (such as standard gaussian noise), the errors are 1.8e-5 (sequential) and 1.5e-5 (parallel).

Speed-up

We tested 3 implementations of the algorithm with the same 100000-element input (code):

  1. Sequential in PyTorch on CPU (as in REINFORCE) (Intel Xeon CPU, DGX-1)
  2. Sequential in C++ on CPU (Intel Xeon CPU, DGX-1)
  3. Parallel in CUDA (NVIDIA P-100, DGX-1)

The observed speed-ups are as follows:

  • PyTorch to C++: 387 times
  • PyTorch to CUDA: 36573 times
  • C++ to CUDA: 94 times

Ops-Space-Time Complexity

Assumptions:

  • A fused operation of raising gamma to a power, multiplying the result by x, and adding y is counted as a single fused operation;
  • N is a power of two. When it isn't, the parallel algorithm's complexity is the same as with N equal to the next power of two.

Under these assumptions, the sequential algorithm takes N operations and N time steps to complete. The parallel algorithm takes 0.5 * N * log2 N operations and can be completed in log2 N time steps if the parallelism is unrestricted.

Both algorithms can be performed in-place; hence their space complexity is O(1).

In Other Frameworks

PyTorch

As of the time of writing, PyTorch does not provide discounted cumsum functionality via the API. PyTorch RL code samples (e.g., REINFORCE) suggest computing returns in a loop over reward items. Since most RL algorithms do not require differentiating through returns, many code samples resort to using SciPy function listed below.

TensorFlow

TensorFlow API provides tf.scan API, which can be supplied with an appropriate lambda function implementing the formula above. Under the hood, however, tf.scan implement the traditional sequential algorithm.

SciPy

SciPy provides a scipy.signal.lfilter function for computing IIR filter response using the sequential algorithm, which can be used for the task at hand, as suggested in this StackOverflow response.

Citation

To cite this repository, use the following BibTeX:

@misc{obukhov2021torchdiscountedcumsum,
  author={Anton Obukhov},
  year=2021,
  title={Fast discounted cumulative sums in PyTorch},
  url={https://github.com/toshas/torch-discounted-cumsum}
}
Owner
Daniel Povey
Daniel Povey
Pytorch bindings for Fortran

Pytorch bindings for Fortran

Dmitry Alexeev 46 Dec 29, 2022
OptNet: Differentiable Optimization as a Layer in Neural Networks

OptNet: Differentiable Optimization as a Layer in Neural Networks This repository is by Brandon Amos and J. Zico Kolter and contains the PyTorch sourc

CMU Locus Lab 428 Dec 24, 2022
Pytorch implementation of Distributed Proximal Policy Optimization

Pytorch-DPPO Pytorch implementation of Distributed Proximal Policy Optimization: https://arxiv.org/abs/1707.02286 Using PPO with clip loss (from https

Alexis David Jacq 164 Jan 05, 2023
On the Variance of the Adaptive Learning Rate and Beyond

RAdam On the Variance of the Adaptive Learning Rate and Beyond We are in an early-release beta. Expect some adventures and rough edges. Table of Conte

Liyuan Liu 2.5k Dec 27, 2022
A pure Python implementation of Compact Bilinear Pooling and Count Sketch for PyTorch.

Compact Bilinear Pooling for PyTorch. This repository has a pure Python implementation of Compact Bilinear Pooling and Count Sketch for PyTorch. This

Grégoire Payen de La Garanderie 234 Dec 07, 2022
A Pytorch Implementation for Compact Bilinear Pooling.

CompactBilinearPooling-Pytorch A Pytorch Implementation for Compact Bilinear Pooling. Adapted from tensorflow_compact_bilinear_pooling Prerequisites I

169 Dec 23, 2022
S3-plugin is a high performance PyTorch dataset library to efficiently access datasets stored in S3 buckets.

S3-plugin is a high performance PyTorch dataset library to efficiently access datasets stored in S3 buckets.

Amazon Web Services 138 Jan 03, 2023
An implementation of Performer, a linear attention-based transformer, in Pytorch

Performer - Pytorch An implementation of Performer, a linear attention-based transformer variant with a Fast Attention Via positive Orthogonal Random

Phil Wang 900 Dec 22, 2022
PyTorch wrappers for using your model in audacity!

PyTorch wrappers for using your model in audacity!

130 Dec 14, 2022
Tez is a super-simple and lightweight Trainer for PyTorch. It also comes with many utils that you can use to tackle over 90% of deep learning projects in PyTorch.

Tez: a simple pytorch trainer NOTE: Currently, we are not accepting any pull requests! All PRs will be closed. If you want a feature or something does

abhishek thakur 1.1k Jan 04, 2023
lookahead optimizer (Lookahead Optimizer: k steps forward, 1 step back) for pytorch

lookahead optimizer for pytorch PyTorch implement of Lookahead Optimizer: k steps forward, 1 step back Usage: base_opt = torch.optim.Adam(model.parame

Liam 318 Dec 09, 2022
Fast Discounted Cumulative Sums in PyTorch

TODO: update this README! Fast Discounted Cumulative Sums in PyTorch This repository implements an efficient parallel algorithm for the computation of

Daniel Povey 7 Feb 17, 2022
Unofficial PyTorch implementation of DeepMind's Perceiver IO with PyTorch Lightning scripts for distributed training

Unofficial PyTorch implementation of DeepMind's Perceiver IO with PyTorch Lightning scripts for distributed training

Martin Krasser 251 Dec 25, 2022
TorchShard is a lightweight engine for slicing a PyTorch tensor into parallel shards

TorchShard is a lightweight engine for slicing a PyTorch tensor into parallel shards. It can reduce GPU memory and scale up the training when the model has massive linear layers (e.g., ViT, BERT and

Kaiyu Yue 275 Nov 22, 2022
Differentiable SDE solvers with GPU support and efficient sensitivity analysis.

PyTorch Implementation of Differentiable SDE Solvers This library provides stochastic differential equation (SDE) solvers with GPU support and efficie

Google Research 1.2k Jan 04, 2023
A lightweight wrapper for PyTorch that provides a simple declarative API for context switching between devices, distributed modes, mixed-precision, and PyTorch extensions.

A lightweight wrapper for PyTorch that provides a simple declarative API for context switching between devices, distributed modes, mixed-precision, and PyTorch extensions.

Fidelity Investments 56 Sep 13, 2022
PyTorch to TensorFlow Lite converter

PyTorch to TensorFlow Lite converter

Omer Ferhat Sarioglu 140 Dec 13, 2022
PyTorch implementation of Glow, Generative Flow with Invertible 1x1 Convolutions

glow-pytorch PyTorch implementation of Glow, Generative Flow with Invertible 1x1 Convolutions

Kim Seonghyeon 433 Dec 27, 2022
PyTorch toolkit for biomedical imaging

farabio is a minimal PyTorch toolkit for out-of-the-box deep learning support in biomedical imaging. For further information, see Wikis and Docs.

San Askaruly 47 Dec 28, 2022
A code copied from google-research which named motion-imitation was rewrited with PyTorch

motor-system Introduction A code copied from google-research which named motion-imitation was rewrited with PyTorch. More details can get from this pr

NewEra 6 Jan 08, 2022