pytorch/examples

A question about weight initialization of embedding layer.

Open

#595 aberto em 18 de jul. de 2019

Ver no GitHub
 (1 comment) (0 reactions) (0 assignees)Python (9.429 forks)batch import
good first issue

Métricas do repositório

Stars
 (21.634 stars)
Métricas de merge de PR
 (Nenhuma PRs mesclada em 30d)

Description

I have a question about the weight intialization of embedding layer.

In source code of PyTorch, the weight of embedding layer is initialized by N(0, 1). In this code, the weight of embedding layer is intialized by uniform.

I trained the model two times with different initialization method, and I found that the default initialization makes the model converges too slow.

So, why the embedding is initialized by N(0, 1), which seems not a good start point?

Guia do colaborador