Fairseq: BART training time

Created on 19 Dec 2019  路  1Comment  路  Source: pytorch/fairseq

May I know how much time BART pre-training took in which GPU configuration? I can see in the paper its written 500K steps with batch size 8k but I want to know the time it took. Many thanks.

question
Source
picture sunilitggu

Most helpful comment

The time can depend on the type and numbers of gpus. We trained for around 11-12 days on 256 gpus.

picture ngoyal2707 on 19 Dec 2019
😄3 👀1 👍1

>All comments

The time can depend on the type and numbers of gpus. We trained for around 11-12 days on 256 gpus.

ngoyal2707 picture ngoyal2707 on 19 Dec 2019
😄3 👀1 👍1
Was this page helpful?
0 / 5 - 0 ratings

Related issues

jmatak picture jmatak  路  3Comments
jordiae picture jordiae  路  3Comments
galphag picture galphag  路  3Comments
ajesujoba picture ajesujoba  路  3Comments
kyquang97 picture kyquang97  路  3Comments