nickluijtgaarden
commented
5 years ago Hi!
Decoding seems to have some issues at beam > 1, because when summaries have
20 sentences, the reranking needs to do too many calculations and just gets stuck.
My solution was too only take the first 20 sentences for the summary and skip the last ones when decoding. On holiday right now, but I can share my code later if you want.
Kind regards,
Nick
Op di 30 jul. 2019 09:45 schreef Timo Wang notifications@github.com:
Hi! I am trying to summarize some text files (not from the cnn daily mail dataset) using your decode_full_model.py. When the number of files is not very big, i.e. around 1000, it was working perfectly fine. However, because I have in total around 1 million files to summarize, the decoding process got stuck at the 1320th file. I tried to restart the decode multiple times and each time it was stuck at the same file. I am wondering what could cause it. I did not modify the code but instead pre process my files so that they have the same structures as the one suggested on the website.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ChenRocks/fast_abs_rl/issues/57?email_source=notifications&email_token=AEHVK6NYJTTB5KH4YXOAXC3QCBHXRA5CNFSM4IH5EXK2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HCKIPHA, or mute the thread https://github.com/notifications/unsubscribe-auth/AEHVK6JFTKYEJP5IIBBLZK3QCBHXRANCNFSM4IH5EXKQ .
timoderbeste
ChenRocks
Hi! I am trying to summarize some text files (not from the cnn daily mail dataset) using your decode_full_model.py. When the number of files is not very big, i.e. around 1000, it was working perfectly fine. However, because I have in total around 1 million files to summarize, the decoding process got stuck at the 1320th file. I tried to restart the decode multiple times and each time it was stuck at the same file. You can see the outputs from the screenshot below.
I am wondering what could cause it. I did not modify the code but instead pre process my files so that they have the same structures as the one suggested on the website.