Merge requests · Administrator / metaseq

[scripts] Convert resharded MP checkpoints to unflattened.

!60 · created May 08, 2022 by Administrator cla signed
- MERGED
- 30
updated Jun 21, 2022
Add missing Namespace import

!126 · created May 27, 2022 by Administrator cla signed
- MERGED
- 3
updated Jun 20, 2022
Support document level attention while training model parallel models 0 of 4 checklist items completed

!104 · created May 17, 2022 by Administrator cla signed
- MERGED
- 22
updated Jun 20, 2022
[api] Simplify launch config.

!157 · created Jun 17, 2022 by Administrator cla signed
- MERGED
- 4
updated Jun 18, 2022
Fix Beam Search

!156 · created Jun 16, 2022 by Administrator cla signed
- MERGED
- 2
updated Jun 17, 2022
fix off by one in API 0 of 4 checklist items completed

!145 · created Jun 13, 2022 by Administrator cla signed
- MERGED
- 5
updated Jun 17, 2022
Fix progress bar exception when both tensorboard and wandb are turned on

!148 · created Jun 13, 2022 by Administrator cla signed
- MERGED
- 2
updated Jun 13, 2022
BF16 support

!30 · created May 04, 2022 by Administrator cla signed
- MERGED
- 20
updated Jun 11, 2022
Add a 2nd md5sum check after consolidating the shards

!141 · created Jun 06, 2022 by Administrator cla signed
- MERGED
- 2
updated Jun 06, 2022
Add instructions to download the GPT2 asset files

!132 · created Jun 01, 2022 by Administrator cla signed
- MERGED
- 2
updated Jun 06, 2022
md5sum of OPT-175B shards

!138 · created Jun 04, 2022 by Administrator cla signed
- MERGED
- 2
updated Jun 04, 2022
Add eos mode for src tgt dataset

!107 · created May 19, 2022 by Administrator cla signed
- MERGED
- 11
updated Jun 02, 2022
Fix errors when build docker 0 of 4 checklist items completed

!130 · created Jun 01, 2022 by Administrator cla signed
- MERGED
- 2
updated Jun 02, 2022
Progress bar cleanup

!96 · created May 12, 2022 by Administrator cla signed
- MERGED
- 9
updated May 31, 2022
Pin omegaconf to 2.1.1, tensorboard to 2.8.0

!127 · created May 28, 2022 by Administrator cla signed
- MERGED
- 2
updated May 30, 2022
Add Dockerfile

!120 · created May 23, 2022 by Administrator cla signed
- MERGED
- 3
updated May 24, 2022
Add Example Proportional Sampling Maximum

!109 · created May 20, 2022 by Administrator cla signed
- MERGED
- 7
updated May 24, 2022
Include diff with black check

!116 · created May 22, 2022 by Administrator cla signed
- MERGED
- 3
updated May 23, 2022
Dynamic loss scaler changes for 66b

!115 · created May 22, 2022 by Administrator cla signed
- MERGED
- 5
updated May 22, 2022
Allow downloading to a target directory

!113 · created May 20, 2022 by Administrator cla signed
- MERGED
- 2
updated May 20, 2022