pskrunner14
commented
6 years ago No @reeuq v0.1.4 does not support initializing embedding layers with pre-trained embeddings as that feature was added in v0.1.6. However if you want to manually initialize the embedding layer, you could use:
self.embedding = nn.Embedding(vocab_size, hidden_size)
self.embedding.weight = nn.Parameter(embedding_matrix)
In the v0.1.4，does it support the use of pre-trained word vectors to initialize the embedding matrix? If so, in the initialization method of class EncoderRNN, when we create the nn.Embedding object, we don't provide the pre-trained word vectors, and how to use the pre-trained word vectors in the embedding layer?