pytorch-wordemb

June 20, 2019 ยท View on GitHub

Load pretrained word embeddings (word2vec, glove format) into torch.FloatTensor for PyTorch

Install

PyTorch required.

pip install torchwordemb

Usage

import torch
import torchwordemb

torchwordemb.load_word2vec_bin(path)

read word2vec binary-format model from path.

returns (vocab, vec)

  • vocab is a dict mapping a word to its index.
  • vec is a torch.FloatTensor of size V x D, where V is the vocabulary size and D is the dimension of word2vec.
vocab, vec = torchwordemb.load_word2vec_bin("/path/to/word2vec/model.bin")
print(vec.size())
print(vec[ w2v.vocab["apple"] ] )

torchwordemb.load_word2vec_text(path)

read word2vec text-format model from path.

torchwordemb.load_glove_text(path)

read GloVe text-format model from path.