Top fairseq issues

A rookie question of RoBERTa finetuning on RACE dataset

bug needs triage
14H034160212 picture 14H034160212·28 Mar 2020·2 Comments

Loss increase with RoBERTa pretraining despite large batch size

question
iyor picture iyor·25 Mar 2020·6 Comments

Best way to implement BPE dropout?

needs triage question
shaywh picture shaywh·23 Mar 2020·5 Comments

alignment indices out of range for sequence of tokens

bug
lhk picture lhk·22 Mar 2020·6 Comments

prepare-iwslt14.sh: chars are replaced with @. Why?

needs triage question
leocaprio picture leocaprio·20 Mar 2020·1 Comments

mBART with CCMatrix

enhancement help wanted needs triage
JohannesTK picture JohannesTK·17 Mar 2020·1 Comments

Fairseq-generate not reading max_positions from --max-len-a

bug needs triage
zeeshansayyed picture zeeshansayyed·17 Mar 2020·5 Comments

speed up beam search by ~2x

enhancement help wanted
yuyan2do picture yuyan2do·17 Mar 2020·1 Comments
👍9 🎉3 👀2

programs hangs in distributed_utils.all_gather_list (reopen)

question
aronla picture aronla·11 Mar 2020·10 Comments

errors trying to decode with mbart model

bug needs triage
mjpost picture mjpost·28 Feb 2020·3 Comments
 Previous issues
Next issues