diff options
author | Marcin Junczys-Dowmunt <Marcin.JunczysDowmunt@microsoft.com> | 2020-01-11 10:05:08 +0300 |
---|---|---|
committer | Marcin Junczys-Dowmunt <Marcin.JunczysDowmunt@microsoft.com> | 2020-01-11 10:05:08 +0300 |
commit | 6889c652851afcb3070bb40dc367809fe0b8d00f (patch) | |
tree | a121be97485710b09d53d91d32dfe437b916ef7a | |
parent | 09698d9e47f9145c0cca6bb1fd5ebe787b7bb4e7 (diff) |
update example for student trainingmjd/v1.9
-rwxr-xr-x | wngt2019/train-student.sh | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/wngt2019/train-student.sh b/wngt2019/train-student.sh index 6d60752..b396d67 100755 --- a/wngt2019/train-student.sh +++ b/wngt2019/train-student.sh @@ -45,9 +45,10 @@ then --valid-metrics bleu-detok ce-mean-words --valid-sets data/valid.{en,de} --quiet-translation \ --log model/train.log --valid-log model/valid.log \ --overwrite --keep-best --seed 1234 --exponential-smoothing \ - --transformer-dropout 0.1 --label-smoothing 0 \ + --transformer-dropout 0.0 --label-smoothing 0 \ --transformer-decoder-autoreg rnn --dec-cell ssru \ --transformer-tied-layers 1 1 1 1 1 1 --dec-depth $DEC_DEPTH \ + --optimizer-params 0.9 0.98 1e-09 0.05 --beam-size 1 fi |