Pytorch cuda extension of grid_sample1d

Overview

Grid Sample 1d

pytorch cuda extension of grid sample 1d. Since pytorch only supports grid sample 2d/3d, I extend the 1d version for efficiency. The forward pass is 2~3x faster than pytorch grid sample.

setup

  • Pytorch == 1.7.1
  • CUDA == 10.1

Other versions of pytorch or cuda may work but I haven't test.

you can choose to manually build it or use JIT

Build

python setup.py install

JIT

comment import grid_sample1d_cuda as grid_sample1d in op.py

uncomment

grid_sample1d = load(
    'grid_sample1d_cuda', ['grid_sample1d_cuda.cpp', 'grid_sample1d_cuda_kernel.cu'], verbose=True)

in op.py

Usage

import torch
from grid_sample1d import GridSample1d

grid_sample1d = GridSample1d(padding_mode=True, align_corners=True)
N = 16
C = 256
L_in = 64
L_out = 128
input = torch.randn((N, C, L_in)).cuda()
grids = torch.randn((N, L_out)).cuda()
output = grid_sample1d(input, grids)

Options are

  • padding_mode: True for border padding, False for zero padding
  • align_corners: same with align_corners in torch.nn.functional.grid_sample

difference

In forward pass, calculation on the channel dim C is parallel, which is serial in torch.nn.functional.grid_sample. Parallel calculation on C may cause round off error in backward. But for now, I found it doesn't influence the forward pass.

Test

Accuracy Test

Since grid sample 1d is a special case of grid sample 2d in most cases (not true when padding_mode & align_corners are both False). I test the accuracy of the implemented grid sample based on torch.nn.functional.grid_sample.

import torch
import torch.nn.functional as F


def gridsample1d_by2d(input, grid, padding_mode, align_corners):
    shape = grid.shape
    input = input.unsqueeze(-1)  # batch_size * C * L_in * 1
    grid = grid.unsqueeze(1)  # batch_size * 1 * L_out
    grid = torch.stack([-torch.ones_like(grid), grid], dim=-1)
    z = F.grid_sample(input, grid, padding_mode=padding_mode, align_corners=align_corners)
    C = input.shape[1]
    out_shape = [shape[0], C, shape[1]]
    z = z.view(*out_shape)  # batch_size * C * L_out
    return z

It is recommended to test on your computer because I only test it on CUDA 10.1 GTX 1080Ti

python test/acc_benchmark.py

Both the forward and the backward results are identical except for align_corners=True, padding_mode=False. It may be caused by round off error when we sum series float numbers in different orders.

Deterministic Test

It is very important to do deterministic test since the associative law is no more applied for the calculation of float numbers on computers.

python test/check_deterministic.py

Note

When padding_mode & align_corners are both False, we cannot regard grid sample 1d as a special case of grid sample 2d in pytorch. I have checked the cuda kernel of grid_sample in Pytorch. When padding_mode & align_corners are both False, the output of torch.nn.functional.grid_sample will be half of the expected. Hope it can be fixed one day.

CPU support

Too lazy to support

speed & memory cost

Here are the speed test results on different size of input

references

Owner
lyricpoem
lyricpoem
PyTorch Implementation of VAENAR-TTS: Variational Auto-Encoder based Non-AutoRegressive Text-to-Speech Synthesis.

VAENAR-TTS - PyTorch Implementation PyTorch Implementation of VAENAR-TTS: Variational Auto-Encoder based Non-AutoRegressive Text-to-Speech Synthesis.

Keon Lee 67 Nov 14, 2022
GestureSSD CBAM - A gesture recognition web system based on SSD and CBAM, using pytorch, flask and node.js

GestureSSD_CBAM A gesture recognition web system based on SSD and CBAM, using pytorch, flask and node.js SSD implementation is based on https://github

xue_senhua1999 2 Jan 06, 2022
Small utility to demangle Nim symbols in callgrind files

nim_callgrind A small utility to demangle Nim symbols from callgrind files. Usage Run your (Nim) program with something like this: valgrind --tool=cal

kraptor 3 Feb 15, 2022
Tutoriais publicados nas nossas redes sociais para obtenção de dados, análises simples e outras tarefas relevantes no mercado financeiro.

Tutoriais Públicos Tutoriais publicados nas nossas redes sociais para obtenção de dados, análises simples e outras tarefas relevantes no mercado finan

Trading com Dados 68 Oct 15, 2022
Differentiable Wavetable Synthesis

Differentiable Wavetable Synthesis

4 Feb 11, 2022
AdamW optimizer and cosine learning rate annealing with restarts

AdamW optimizer and cosine learning rate annealing with restarts This repository contains an implementation of AdamW optimization algorithm and cosine

Maksym Pyrozhok 133 Dec 20, 2022
Python code for the paper How to scale hyperparameters for quickshift image segmentation

How to scale hyperparameters for quickshift image segmentation Python code for the paper How to scale hyperparameters for quickshift image segmentatio

0 Jan 25, 2022
PyTorch implementation of PSPNet

PSPNet with PyTorch Unofficial implementation of "Pyramid Scene Parsing Network" (https://arxiv.org/abs/1612.01105). This repository is just for caffe

Kazuto Nakashima 52 Nov 16, 2022
Pytorch Implementation of "Contrastive Representation Learning for Exemplar-Guided Paraphrase Generation"

CRL_EGPG Pytorch Implementation of Contrastive Representation Learning for Exemplar-Guided Paraphrase Generation We use contrastive loss implemented b

YHR 25 Nov 14, 2022
Implement Decoupled Neural Interfaces using Synthetic Gradients in Pytorch

disclaimer: this code is modified from pytorch-tutorial Image classification with synthetic gradient in Pytorch I implement the Decoupled Neural Inter

Andrew 114 Dec 22, 2022
Pytorch Implementation of Auto-Compressing Subset Pruning for Semantic Image Segmentation

Pytorch Implementation of Auto-Compressing Subset Pruning for Semantic Image Segmentation Introduction ACoSP is an online pruning algorithm that compr

Merantix 8 Dec 07, 2022
I-BERT: Integer-only BERT Quantization

I-BERT: Integer-only BERT Quantization HuggingFace Implementation I-BERT is also available in the master branch of HuggingFace! Visit the following li

Sehoon Kim 139 Dec 27, 2022
A Re-implementation of the paper "A Deep Learning Framework for Character Motion Synthesis and Editing"

What is This This is a simple re-implementation of the paper "A Deep Learning Framework for Character Motion Synthesis and Editing"(1). Only Sections

102 Dec 14, 2022
Revisiting, benchmarking, and refining Heterogeneous Graph Neural Networks.

Heterogeneous Graph Benchmark Revisiting, benchmarking, and refining Heterogeneous Graph Neural Networks. Roadmap We organize our repo by task, and on

THUDM 176 Dec 17, 2022
Autotype on websites that have copy-paste disabled like Moodle, HackerEarth contest etc.

Autotype A quick and small python script that helps you autotype on websites that have copy paste disabled like Moodle, HackerEarth contests etc as it

Tushar 32 Nov 03, 2022
This repository provides data for the VAW dataset as described in the CVPR 2021 paper titled "Learning to Predict Visual Attributes in the Wild"

Visual Attributes in the Wild (VAW) This repository provides data for the VAW dataset as described in the CVPR 2021 Paper: Learning to Predict Visual

Adobe Research 36 Dec 30, 2022
Creating Multi Task Models With Keras

Creating Multi Task Models With Keras About The Project! I used the keras and Tensorflow Library, To build a Deep Learning Neural Network to Creating

Srajan Chourasia 4 Nov 28, 2022
Binary Stochastic Neurons in PyTorch

Binary Stochastic Neurons in PyTorch http://r2rt.com/binary-stochastic-neurons-in-tensorflow.html https://github.com/pytorch/examples/tree/master/mnis

Onur Kaplan 54 Nov 21, 2022
Code and datasets for the paper "Combining Events and Frames using Recurrent Asynchronous Multimodal Networks for Monocular Depth Prediction" (RA-L, 2021)

Combining Events and Frames using Recurrent Asynchronous Multimodal Networks for Monocular Depth Prediction This is the code for the paper Combining E

Robotics and Perception Group 69 Dec 26, 2022
Structure Information is the Key: Self-Attention RoI Feature Extractor in 3D Object Detection

Structure Information is the Key: Self-Attention RoI Feature Extractor in 3D Object Detection abstract:Unlike 2D object detection where all RoI featur

DK. Zhang 2 Oct 07, 2022