https://github.com/magnusja/conve

Implementation of ConvE proposed by Dettmers et al. in Convolutional 2D Knowledge Graph Embeddings.
https://github.com/magnusja/conve

conve convolutional-neural-networks deep-learning knowledge-embedding knowledge-graph machine-learning network-embedding neural-network

Last synced: 10 months ago
JSON representation

Implementation of ConvE proposed by Dettmers et al. in Convolutional 2D Knowledge Graph Embeddings.

• Host: GitHub
• URL: https://github.com/magnusja/conve
• Owner: magnusja
• License: apache-2.0
• Created: 2017-07-13T05:53:45.000Z (almost 9 years ago)
• Default Branch: master
• Last Pushed: 2018-05-24T15:04:33.000Z (almost 8 years ago)
• Last Synced: 2025-04-20T07:58:19.206Z (about 1 year ago)
• Topics: conve, convolutional-neural-networks, deep-learning, knowledge-embedding, knowledge-graph, machine-learning, network-embedding, neural-network
• Language: Python
• Homepage:
• Size: 36.1 KB
• Stars: 15
• Watchers: 5
• Forks: 7
• Open Issues: 1
• Metadata Files:
  • Readme: Readme.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# ConvE

Implementation of ConvE proposed by Dettmers et al. in [Convolutional 2D Knowledge Graph Embeddings](https://arxiv.org/abs/1707.01476). You can find the official repository with knowledge graph datasets [here](https://github.com/TimDettmers/ConvE).

Implementation uses [PyTorch](http://pytorch.org/).

## Usage

### Preprocessing

```

usage: preprocess.py [-h] {train,valid} ...

Preprocess knowledge graph csv train/valid (test) data.

positional arguments:

  {train,valid}  mode

    train        Preprocess a training set

    valid        Preprocess a valid or test set

optional arguments:

  -h, --help     show this help message and exit

```

#### Training set

```

python preprocess.py train ../train.tsv

```

#### Validation set

```

python preprocess.py valid ../train.pkl ../valid.tsv

```

### Training

```

python train.py ../train.pkl ../valid.pkl

```

```

usage: train.py [-h] [--name NAME] [--batch-size BATCH_SIZE] [--epochs EPOCHS]

                [--label-smooth LABEL_SMOOTH]

                train_path valid_path

Train ConvE with PyTorch.

positional arguments:

  train_path            Path to training .pkl produced by preprocess.py

  valid_path            Path to valid/test .pkl produced by preprocess.py

optional arguments:

  -h, --help            show this help message and exit

  --name NAME           name of the model, used to create a subfolder to save

                        checkpoints

  --batch-size BATCH_SIZE

  --epochs EPOCHS

  --label-smooth LABEL_SMOOTH

```