Commits · e7f0cdee75001907e2b6cdce9a41ac0ee48d8181 · OpenDAS / Megatron-LM

23 Feb, 2022 2 commits
- renamed reduce_gradients -> reduce_grads [ matches gather_params ] · e7f0cdee
  Lawrence McAfee authored Feb 23, 2022
  
  e7f0cdee
- fixed param_world_shard bug. · 4b843668
  Lawrence McAfee authored Feb 23, 2022
  
  4b843668
22 Feb, 2022 4 commits
- debugging; localized issue to gather_params() · c13c0a3e
  Lawrence McAfee authored Feb 22, 2022
  
  c13c0a3e
- more cleanup of main params/grads · 9b7854e4
  Lawrence McAfee authored Feb 22, 2022
  
  9b7854e4
- removed zero-size optimizer group shards. · 291592e4
  Lawrence McAfee authored Feb 22, 2022
  
  291592e4
- set main_param_shard==None if local group_size==0 [ e.g., word embedding params · 23f9238d
  Lawrence McAfee authored Feb 22, 2022
  
  23f9238d
18 Feb, 2022 5 commits
- small fixes; training & learning! · ac5ef637
  Lawrence McAfee authored Feb 18, 2022
  
  ac5ef637
- copy model params to main params at very start · a4f41882
  Lawrence McAfee authored Feb 18, 2022
  
  a4f41882
- some cleanup · ae6cd9b5
  Lawrence McAfee authored Feb 18, 2022
  
  ae6cd9b5
- modularized reduce_gradients, gather params; trainin runs, but loss==nan · e5db0fda
  Lawrence McAfee authored Feb 18, 2022
  
  e5db0fda
- fixed clip_grad_norm bug · a7782b21
  Lawrence McAfee authored Feb 18, 2022
  
  a7782b21
16 Feb, 2022 1 commit
- small changes · d1f5776a
  Lawrence McAfee authored Feb 15, 2022
  
  d1f5776a
15 Feb, 2022 7 commits
- reduce scatter working · ff4c4273
  Lawrence McAfee authored Feb 15, 2022
  
  ff4c4273
- added 'all' world shards to gbuf map · bf64c85c
  Lawrence McAfee authored Feb 15, 2022
  
  bf64c85c
- updated zero_grad · eaa0c1df
  Lawrence McAfee authored Feb 15, 2022
  
  eaa0c1df
- allocated main params/grads · 33626179
  Lawrence McAfee authored Feb 15, 2022
  
  33626179
- created optimizer group shards · 525a8351
  Lawrence McAfee authored Feb 15, 2022
  
  525a8351
- modularized shard indexing · f498a6fe
  Lawrence McAfee authored Feb 15, 2022
  
  f498a6fe
- wip; switching to grad-buffer-centric design · cb6f96b6
  Lawrence McAfee authored Feb 15, 2022
  
  cb6f96b6
14 Feb, 2022 8 commits
- todo; align shards with model's contiguous buffer · a3f3c3ad
  Lawrence McAfee authored Feb 14, 2022
  
  a3f3c3ad
- copying model grad slices to main grad · 3f0bc681
  Lawrence McAfee authored Feb 14, 2022
  
  3f0bc681
- fix zero_grad; set_to_none = False · 6875dff5
  Lawrence McAfee authored Feb 14, 2022
  
  6875dff5
- tweaked slice index naming convention · 1215c420
  Lawrence McAfee authored Feb 14, 2022
  
  1215c420
- map param to originating virtual model; eventually move this to constructor · c5f93269
  Lawrence McAfee authored Feb 14, 2022
  
  c5f93269
- included original param index in map · 3ded2425
  Lawrence McAfee authored Feb 14, 2022
  
  3ded2425
- built local shard param index map · a74e245c
  Lawrence McAfee authored Feb 14, 2022
  
  a74e245c
- collect param offsets for contiguous grad buffer · f7232502
  Lawrence McAfee authored Feb 14, 2022
  
  f7232502
11 Feb, 2022 2 commits
- bit more progress · 5706ba42
  Lawrence McAfee authored Feb 11, 2022
  
  5706ba42
- studied float16 optimizer; more updates · f48e1f29
  Lawrence McAfee authored Feb 11, 2022
  
  f48e1f29
10 Feb, 2022 2 commits
- more work on Float16DistributedOptimizer · 49cca4d9
  Lawrence McAfee authored Feb 10, 2022
  
  49cca4d9
- working on Float16DistributedOptimizer · 329fe582
  Lawrence McAfee authored Feb 10, 2022
  
  329fe582
09 Feb, 2022 1 commit
- feb 9 alpha · 7dc8c475
  Lawrence McAfee authored Feb 09, 2022
  
  7dc8c475
29 Jan, 2022 2 commits
- typo fix · 2b628f96
  Vijay Korthikanti authored Jan 28, 2022
  
  2b628f96
- Revert incorrect fix. · cd499559
  Jared Casper authored Jan 28, 2022
  
  cd499559
28 Jan, 2022 2 commits
- more naming cleanup · 641408f5
  Vijay Korthikanti authored Jan 28, 2022
  
  641408f5
- changing class name AnnealingLR to OptimizerParamScheduler · 04ecc834
  Vijay Korthikanti authored Jan 28, 2022
  
  04ecc834
27 Jan, 2022 1 commit
- address review comments · 53931b8b
  Vijay Korthikanti authored Jan 27, 2022
  
  53931b8b
26 Jan, 2022 2 commits
- further clarified viewless tensor comment in transformer.py · d16e2a24
  Lawrence McAfee authored Jan 26, 2022
  
  d16e2a24
- address review comments · 8acbbe25
  Vijay Korthikanti authored Jan 26, 2022
  
  8acbbe25
25 Jan, 2022 1 commit
- limit 'make_viewless_tensor()' to case of micro_batch_size == 1; added comment · 24369dd6
  Lawrence McAfee authored Jan 25, 2022
  
  24369dd6