Mask prob

["--model","xfmr","--num-layers","3","--max-seq-len","512","--name","source-layer3","--train-data","saved.unsup/processed/train.tar","--valid-data","saved.unsup/processed/val.tar","--test-data","saved.unsup/processed/test.tar","--vocab-path","saved.unsup/vocab.model","--inst-prob-vocab-path","saved.unsup/vocab.inst","--model","xfmr_enc","--program-input-rep","source","--datastore-data","saved.poj-512-480/raw/poj-512-480-train.tar"]

["--model","xfmr","--num-layers","3","--max-seq-len","512","--name","source-layer3-mask-0.5","--train-data","saved.unsup/processed/train.tar","--valid-data","saved.unsup/processed/val.tar","--test-data","saved.unsup/processed/test.tar","--vocab-path","saved.unsup/vocab.model","--inst-prob-vocab-path","saved.unsup/vocab.inst","--model","xfmr_enc","--program-input-rep","source","--datastore-data","saved.poj-512-480/raw/poj-512-480-train.tar","--mask-prob","0.5"]

["--model","xfmr","--num-layers","3","--max-seq-len","512","--name","random","--train-data","saved.unsup/processed/train.tar","--valid-data","saved.unsup/processed/val.tar","--test-data","saved.unsup/processed/test.tar","--vocab-path","saved.unsup/vocab.model","--inst-prob-vocab-path","saved.unsup/vocab.inst","--model","xfmr_enc","--program-input-rep","source","--datastore-data","saved.poj-512-480/raw/poj-512-480-train.tar"]

{"remote":"git@github.com:qibinc/nncompiler.git","commit":"7c8a76de768e77bb0f6e50b6dfaaee66ff15a411","__typename":"GitInfo"}

{"remote":"git@github.com:qibinc/nncompiler.git","commit":"ae3a97e9fb84196b9d2e175c0a3b1170fb3c2d5d","__typename":"GitInfo"}

{"remote":"git@github.com:qibinc/nncompiler.git","commit":"ae3a97e9fb84196b9d2e175c0a3b1170fb3c2d5d","__typename":"GitInfo"}

23h 6m 18s

22h 22m 39s

2h 42m 19s

0.15

0.5

0.15

source-layer3

source-layer3-mask-0.5

random