Merge requests · Administrator / metaseq

[tboard] Don't run out of letters

!527 · created Nov 22, 2022 by Administrator cla signed
- MERGED
- 2
updated Nov 22, 2022
Add data proportions

!488 · created Nov 04, 2022 by Administrator opt_instruct cla signed
- MERGED
- 9
updated Nov 21, 2022
Update setup.md

!524 · created Nov 18, 2022 by Administrator cla signed
- MERGED
- 2
updated Nov 19, 2022
Log host where creating a directory can fail

!525 · created Nov 18, 2022 by Administrator cla signed
- MERGED
- 2
updated Nov 18, 2022
Truncate normal init

!522 · created Nov 17, 2022 by Administrator cla signed
- MERGED
- 2
updated Nov 17, 2022
Adding relu for seq par

!517 · created Nov 14, 2022 by Administrator cla signed
- MERGED
- 7
updated Nov 16, 2022
Add headers

!514 · created Nov 12, 2022 by Administrator cla signed
- MERGED
- 3
updated Nov 14, 2022
Add a script to reshard FSDP checkpoints 0 of 4 checklist items completed

!459 · created Oct 26, 2022 by Administrator cla signed
- MERGED
- 19
updated Nov 14, 2022
Added dynamic configs (profiling) 0 of 4 checklist items completed

!473 · created Nov 01, 2022 by Administrator cla signed
- MERGED
- 31
updated Nov 11, 2022
[distributed] Custom rendezvous

!510 · created Nov 11, 2022 by Administrator cla signed
- MERGED
- 3
updated Nov 11, 2022
Revert "Added dynamic configs (profiling) (#473)"

!509 · created Nov 11, 2022 by Administrator cla signed
- MERGED
- 3
updated Nov 11, 2022
skip PathManager stuff for NFS save

!497 · created Nov 08, 2022 by Administrator cla signed
- MERGED
- 11
updated Nov 10, 2022
Add args check for seq par

!505 · created Nov 10, 2022 by Administrator cla signed
- MERGED
- 2
updated Nov 10, 2022
Remove gloo pg for now, tcp issue

!504 · created Nov 09, 2022 by Administrator cla signed
- MERGED
- 2
updated Nov 09, 2022
Split to multiple sbatch copy jobs

!502 · created Nov 09, 2022 by Administrator cla signed
- MERGED
- 6
updated Nov 09, 2022
Save before validating, remove best* checkpoint logic

!499 · created Nov 08, 2022 by Administrator cla signed
- MERGED
- 2
updated Nov 09, 2022
Fix verify shards - to fix resuming from checkpoint

!501 · created Nov 08, 2022 by Administrator opt_instruct cla signed
- MERGED
- 4
updated Nov 09, 2022
isolated PR for seq_par with autograd fn to have more control over op…

!479 · created Nov 02, 2022 by Administrator cla signed
- MERGED
- 16
updated Nov 08, 2022
Launch separate sbatch job to copy checkpoints over from scratch to NFS

!494 · created Nov 06, 2022 by Administrator cla signed
- MERGED
- 21
updated Nov 07, 2022
Replace jobid in local checkpoint dir with actual job id

!491 · created Nov 04, 2022 by Administrator cla signed
- MERGED
- 2
updated Nov 05, 2022