Commit cc85d411 authored by Stephen Roller's avatar Stephen Roller Committed by Myle Ott
Browse files

Fix a bug with using GloVe 840B tokens for initialization.

parent 9dcee4c7

fairseq/utils.py

+1 −1
Changes for fairseq/utils.py: 1 added line, 1 removed line.
Original line number Diff line number Diff line
@@ -256,7 +256,7 @@ def parse_embedding(embed_path): 
    with open(embed_path) as f_embed:

        next(f_embed)  # skip header

        for line in f_embed:

            pieces = line.strip().split()

            pieces = line.rstrip().split(" ")

            embed_dict[pieces[0]] = torch.Tensor([float(weight) for weight in pieces[1:]])

    return embed_dict