Skip to main content

Minimal torch/numpy agnostic relational function.

Project description

A Minimal Relational Function

PyTorch/Numpy agnostic function implementing the relational block from "A simple neural network module for relational reasoning".

I wrote this to learn how to use nbdev. I'm pretty sure it's correct but it only implements the core function for using relational networks and none of the other stuff (such as nn.Module classes etc) that Kai included in the pull request.

The original paper can be found here.

Install

pip install relational

How to use

This can be used to implement a relational network in PyTorch. An example would be something like:

from relational.core import relation
import torch
import torch.nn as nn
class SetNet(nn.Module):
    def __init__(self, datadim, n_hidden):
        super(SetNet, self).__init__()
        self.n_hidden = n_hidden
        self.g = nn.Sequential(nn.Linear(datadim*2, n_hidden), 
                               nn.ReLU(),
                               nn.Linear(n_hidden, n_hidden))
        self.f = nn.Sequential(nn.Linear(n_hidden, n_hidden),
                               nn.ReLU(),
                               nn.Linear(n_hidden, n_hidden))

    def forward(self, x):
        n, t, d = x.size()
        x = relation(x, self.g, reduction='mean')
        return self.f(x)
x = torch.randn(4, 8, 16)
setnet = SetNet(x.size(2), 10)
setnet(x).size()
torch.Size([4, 10])

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

relational-0.0.1.tar.gz (10.5 kB view details)

Uploaded Source

Built Distribution

relational-0.0.1-py3-none-any.whl (7.9 kB view details)

Uploaded Python 3

File details

Details for the file relational-0.0.1.tar.gz.

File metadata

  • Download URL: relational-0.0.1.tar.gz
  • Upload date:
  • Size: 10.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.4.1 importlib_metadata/4.3.1 pkginfo/1.7.0 requests/2.25.1 requests-toolbelt/0.9.1 tqdm/4.61.0 CPython/3.9.5

File hashes

Hashes for relational-0.0.1.tar.gz
Algorithm Hash digest
SHA256 7d59d45465b39380ed1b6f4a5c272aa9c5f32fa59cf7c9615eacfa1334be704f
MD5 c477ed4527ad497f25e1f2dda7d37934
BLAKE2b-256 06a1124e6f45d9a88c8168fef64ed54b0d40131eebb075106a18c0420d4d5a27

See more details on using hashes here.

File details

Details for the file relational-0.0.1-py3-none-any.whl.

File metadata

  • Download URL: relational-0.0.1-py3-none-any.whl
  • Upload date:
  • Size: 7.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/3.4.1 importlib_metadata/4.3.1 pkginfo/1.7.0 requests/2.25.1 requests-toolbelt/0.9.1 tqdm/4.61.0 CPython/3.9.5

File hashes

Hashes for relational-0.0.1-py3-none-any.whl
Algorithm Hash digest
SHA256 888003da117b268c1acd13dec8a0d1ad267165994a542f79cd64c28ca3e919f0
MD5 fdf19ec175a05296b6d83dc5a1fe6102
BLAKE2b-256 c808846dd4f44b192b7d473e8d2002b6aa680b387fe3ba640038aec2c09620bb

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page