提交 5e353a87 编写于 作者: 1024的传说's avatar 1024的传说 提交者: Guo Sheng

update_seq2seq_readme (#3783)

上级 a9abd027
...@@ -110,14 +110,18 @@ python infer.py \ ...@@ -110,14 +110,18 @@ python infer.py \
mosesdecoder/scripts/generic/multi-bleu.perl tst2013.vi < infer_output.txt mosesdecoder/scripts/generic/multi-bleu.perl tst2013.vi < infer_output.txt
``` ```
单个模型 beam_size = 10的效果如下 每个模型分别训练了10次,单次取第10个epoch保存的模型进行预测,取beam_size=10。效果如下(为了便于观察,对10次结果按照升序进行了排序)
``` ```
> no attention > no attention
tst2012 BLEU: 10.99 tst2012 BLEU:
tst2013 BLEU: 11.23 [10.75 10.85 10.9 10.94 10.97 11.01 11.01 11.04 11.13 11.4]
tst2013 BLEU:
>with attention [10.71 10.71 10.74 10.76 10.91 10.94 11.02 11.16 11.21 11.44]
tst2012 BLEU: 22.85
tst2013 BLEU: 25.68 > with attention
tst2012 BLEU:
[21.14 22.34 22.54 22.65 22.71 22.71 23.08 23.15 23.3 23.4]
tst2013 BLEU:
[23.41 24.79 25.11 25.12 25.19 25.24 25.39 25.61 25.61 25.63]
``` ```
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册