Commits · c5be45eea204c8a2922f2db4462cb4a60fc5d981 · OpenDAS / Megatron-LM · GitLab

29 Mar, 2022 7 commits
- commented distrib_optimizer.py. · b42c3052
  Lawrence McAfee authored Mar 29, 2022
  
  b42c3052
- partially commented distrib_optimizer.py. · fb3a1345
  Lawrence McAfee authored Mar 29, 2022
  
  fb3a1345
- replace triple single w/ triple double quote. · fe2d623e
  Lawrence McAfee authored Mar 29, 2022
  
  fe2d623e
- renamed: full_ -> model_. · 7ed649ed
  Lawrence McAfee authored Mar 29, 2022
  
  7ed649ed
- commented optimizer.py. · 82e6730c
  Lawrence McAfee authored Mar 29, 2022
  
  82e6730c
- removed debug_base/main/model methods. · 087cbff1
  Lawrence McAfee authored Mar 29, 2022
  
  087cbff1
- fixed shared weight attribute for fp32. · be8de1b3
  Lawrence McAfee authored Mar 29, 2022
  
  be8de1b3
28 Mar, 2022 1 commit
- correctly copying tensor parallel attributes for fp32_groups. · b1781f31
  Lawrence McAfee authored Mar 28, 2022
  
  b1781f31
25 Mar, 2022 7 commits
- debugging fp32 training. · d0925652
  Lawrence McAfee authored Mar 25, 2022
  
  d0925652
- working: checkpoint save/load. · 9f64f5f4
  Lawrence McAfee authored Mar 25, 2022
  
  9f64f5f4
- more cleanup; timing tests vs. 'dupe' dist-opt. · 82491e4b
  Lawrence McAfee authored Mar 24, 2022
  
  82491e4b
- cleanup. · 91f3579e
  Lawrence McAfee authored Mar 24, 2022
  
  91f3579e
- losses match. · 4b9a218e
  Lawrence McAfee authored Mar 24, 2022
  
  4b9a218e
- training; loss not yet checked. · 41ad0c5d
  Lawrence McAfee authored Mar 24, 2022
  
  41ad0c5d
- working: copy grads. · fe3cfd86
  Lawrence McAfee authored Mar 24, 2022
  
  fe3cfd86
24 Mar, 2022 3 commits
- stalling in copy_grads(). · 5e29905f
  Lawrence McAfee authored Mar 24, 2022
  
  5e29905f
- many edits; working towards first draft. · 2c3cb9fc
  Lawrence McAfee authored Mar 24, 2022
  
  2c3cb9fc
- getting started with no-grad-duplication implementation. · 867105c2
  Lawrence McAfee authored Mar 23, 2022
  
  867105c2
23 Mar, 2022 1 commit
- removed mem savings factor. +cleanup. · 11581195
  Lawrence McAfee authored Mar 23, 2022
  
  11581195
22 Mar, 2022 4 commits
- running & saving memory w/ _reduce_scatter_base/_all_gather_base. · 41038d54
  Lawrence McAfee authored Mar 22, 2022
  
  41038d54
- setup code to try _reduce_scatter_base, _all_gather_base. · 64b94f00
  Lawrence McAfee authored Mar 22, 2022
  
  64b94f00
- grad norm 'matches' (not bitwise equal). · 6728a780
  Lawrence McAfee authored Mar 22, 2022
  
  6728a780
- implemented 'get_grad_views_for_grad_norm()'. · c6d20c05
  Lawrence McAfee authored Mar 22, 2022
  
  c6d20c05
18 Mar, 2022 5 commits
- cleanup. · 3d433e8a
  Lawrence McAfee authored Mar 18, 2022
  
  3d433e8a
- using smaller chunk_numel_min. · 19730725
  Lawrence McAfee authored Mar 18, 2022
  
  19730725
- mem savings factor training; need to check loss. · ed90a1b2
  Lawrence McAfee authored Mar 18, 2022
  
  ed90a1b2
- after experiments, before new arg 'distrib-opt-comm-buf-1m'. · 34fc1e37
  Lawrence McAfee authored Mar 18, 2022
  
  34fc1e37
- split reduce_grads/gather_params using gbuf sub-views. · 12d91733
  Lawrence McAfee authored Mar 18, 2022
  
  12d91733
14 Mar, 2022 4 commits
- cleaned distrib_optimizer.py. · 2c1660e7
  Lawrence McAfee authored Mar 14, 2022
  
  2c1660e7
- more specific formatting of model/optim checkpoint paths. · 9c86abd9
  Lawrence McAfee authored Mar 14, 2022
  
  9c86abd9
- optimizer saves list(group), not list(param). · af2b136f
  Lawrence McAfee authored Mar 14, 2022
  
  af2b136f
- started updating checkpointing.py. · 37ca7859
  Lawrence McAfee authored Mar 14, 2022
  
  37ca7859
11 Mar, 2022 1 commit
- guard 'gather_params()' with 'if update_successful:' · f6811e28
  Lawrence McAfee authored Mar 11, 2022
  
  f6811e28
10 Mar, 2022 1 commit
- updated args for allreduce_embeddings · 08ee8ea2
  Lawrence McAfee authored Mar 10, 2022
  
  08ee8ea2
09 Mar, 2022 6 commits
- removed old (pre-gbuf-padding) get_model_grad_buffer_dp_views(). · 7ac342b7
  Lawrence McAfee authored Mar 09, 2022
  
  7ac342b7
- more cleanup · 0528bf77
  Lawrence McAfee authored Mar 09, 2022
  
  0528bf77
- removed faulty (already commented out) gather_model_params param iterator. · 94a90215
  Lawrence McAfee authored Mar 09, 2022
  
  94a90215
- renamed reduce_grads/gather_params -> reduce_model_grads/gather_model_params · 64b9d28c
  Lawrence McAfee authored Mar 09, 2022
  
  64b9d28c
- passing args, timers from train_step to optimizer methods. · d58d1762
  Lawrence McAfee authored Mar 09, 2022
  
  d58d1762
- consolidated reduce_grad's sub-methods (i.e., allreduce_embedding_grads) · 45b364b1
  Lawrence McAfee authored Mar 09, 2022
  
  45b364b1