|
Print the reason why training stops
|
Tolstoyevsky
|
|
5 years ago |
|
Add standalone binaries
|
Myle Ott
|
|
5 years ago |
|
Do distributed init after data loading
|
Myle Ott
|
|
5 years ago |
|
Add code for "Pay Less Attention with Lightweight and Dynamic Convolutions" (#473)
|
Myle Ott
|
|
5 years ago |
|
Print model and number of trained params
|
Myle Ott
|
|
5 years ago |
|
FIX: '--user-dir' on multi-gpu (#449)
|
Davide Caroselli
|
|
5 years ago |
|
Misc fixes
|
Myle Ott
|
|
5 years ago |
|
Merge internal changes (#283)
|
Myle Ott
|
|
5 years ago |
|
Make multiprocessing_train.py work with multi-node setups
|
Myle Ott
|
|
5 years ago |
|
Take a dummy train step under OOM to keep multiprocessing in sync
|
Halil Akin
|
|
5 years ago |
|
Merge small fixes from internal
|
Naman Goyal
|
|
5 years ago |
|
Manually port pull request 385
|
Peng-Jen Chen
|
|
5 years ago |
|
Merge internal changes (#295)
|
Myle Ott
|
|
5 years ago |
|
Switch to DistributedDataParallelC10d and bump version 0.5.0 -> 0.6.0
|
Sergey Edunov
|
|
5 years ago |
|
Add documentation
|
Myle Ott
|
|
5 years ago |
|
Clean up FairseqTask so that it's easier to extend/add new tasks
|
Myle Ott
|
|
5 years ago |
|
dont send dummy batch when reloading from checkpoint
|
Alexei Baevski
|
|
5 years ago |
|
Add training wall time meter
|
Myle Ott
|
|
5 years ago |
|
Warn when using FP16 on pre-Volta GPUs
|
Myle Ott
|
|
5 years ago |
|
Reset gnorm after each epoch
|
Sergey Edunov
|
|
5 years ago |
|
cosine + triangular lr scheduler
|
Alexei Baevski
|
|
5 years ago |
|
add flag that allows keeping optimizer config
|
alexeib
|
|
5 years ago |
|
Transformer lm
|
Alexei Baevski
|
|
5 years ago |
|
Fix interpretation of --max-epoch
|
Myle Ott
|
|
6 years ago |
|
Store full checkpoints instead of symlinking
|
Myle Ott
|
|
6 years ago |
|
Add FairseqTask
|
Myle Ott
|
|
6 years ago |
|
build optimizer only once, otherwise it leaks cuda memory
|
Alexei Baevski
|
|
6 years ago |
|
Small fixes
|
Myle Ott
|
|
6 years ago |
|
Merge validate and val_loss functions (simplify train.py)
|
Myle Ott
|
|
6 years ago |
|
Use symlinks for redundant checkpoints
|
Myle Ott
|
|
6 years ago |