Current model is a seq2seq model with per word attention.
Current baseline performance on different tasks:
Task | Accuracy (%) after 30 epochs
1 | 100
2 | 91.63
3 | 73.48
4 | 100
5 | 99.29
6 | 99.19
7 | 96.06
8 |
9 | 97.27
10 | 99.5
11 | 100
12 | 100
13 | 96.47
14 | 99
15 | 100
16 | 51.2
17 | 62.9
18 | 99.89
19 | 34.17
20 | 99.89