Skip to content

Latest commit

 

History

History
 
 

neural-machine-translation

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

How-to

  1. Run download-preprocess-dataset.ipynb to download dataset and preprocessing.
  2. Run any notebook using Jupyter Notebook.

Accuracy, not sorted

  1. Trainset to train, validation and test set to test
  2. Based on 20 epochs
  3. Accuracy based on word positions
  4. Some results are empty because the models are slow to train, still waiting for the results
  5. Sort from shortest length to longest length and do bucketing from it will improve the accuracy by a lot.
name accuracy
1.basic-seq2seq 0.103391
2.lstm-seq2seq 0.118877
3.gru-seq2seq 0.115032
4.basic-seq2seq-contrib-greedy 0.252812
5.lstm-seq2seq-contrib-greedy 0.330939
6.gru-seq2seq-greedy 0.312779
7.basic-birnn-seq2seq-manual 0.125462
8.lstm-birnn-seq2seq-manual 0.121065
9.gru-birnn-seq2seq-manual 0.119774
10.basic-birnn-seq2seq-greedy 0.274987
11.lstm-birnn-seq2seq-greedy 0.342469
12.gru-birnn-seq2seq-greedy 0.325840
13.basic-seq2seq-luong 0.023959
14.lstm-seq2seq-luong 0.130840
15.gru-seq2seq-luong 0.073492
16.basic-seq2seq-bahdanau 0.132169
17.lstm-seq2seq-bahdanau 0.133821
18.gru-seq2seq-bahdanau 0.140176
19.basic-birnn-seq2seq-bahdanau 0.138824
20.lstm-birnn-seq2seq-bahdanau 0.131571
21.gru-birnn-seq2seq-bahdanau 0.134661
22.basic-birnn-seq2seq-luong 0.074942
23.lstm-birnn-seq2seq-luong 0.132617
24.gru-birnn-seq2seq-luong 0.137604
25.lstm-seq2seq-contrib-greedy-luong 0.455244
26.gru-seq2seq-contrib-greedy-luong 0.081386
27.lstm-seq2seq-contrib-greedy-bahdanau 0.438774
28.gru-seq2seq-contrib-greedy-bahdanau 0.441251
29.lstm-seq2seq-contrib-beam-bahdanau 0.244880
30.gru-seq2seq-contrib-beam-bahdanau 0.222577
31.lstm-birnn-seq2seq-contrib-beam-luong 0.241488
32.gru-birnn-seq2seq-contrib-beam-luong 0.223249
33.lstm-birnn-seq2seq-contrib-luong-bahdanau-beam
34.gru-birnn-seq2seq-contrib-luong-bahdanau-beam
35.bytenet-greedy
36.capsule-lstm-seq2seq-contrib-greedy
37.capsule-gru-seq2seq-contrib-greedy
38.dnc-seq2seq-bahdanau-greedy
39.dnc-seq2seq-luong-greedy
40.lstm-birnn-seq2seq-beam-luongmonotic 0.272306
41.lstm-birnn-seq2seq-beam-bahdanaumonotic 0.263432
42.memory-network-lstm-seq2seq-contrib 0.280245
43.attention-is-all-you-need-beam 0.378041
44.conv-seq2seq 0.337291
45.conv-encoder-lstm-decoder 0.329069
46.dilated-conv-seq2seq 0.331730
47.gru-birnn-seq2seq-greedy-residual 0.343508
48.google-nmt 0.330886
49.bert-transformer-decoder-beam 0.446938
50.xlnet-base-transformer-decoder-beam 0.288339
51.evolved-transformer-tiny 0.497402