GitLab

* [feat] save memory by using bucket buffer only in backward

- this fixes bug #627
- added documentation to clarify the buffer's cost and speed/memory
  tradeoff
- added setup/teardown calls so that the buffer is only allocated
  during the backward pass, saving more memory for forward and stepping
  so that they can be used for things like activations.
- added a unit test that assert the memory is in range.

Comparing with DDP:

  1. buffer size scales with # of FSDP not model size
  2. buffer is only allocated during backward
  3. buffer is used for small tensors only to reduce overhead
  4. overlapping of compute-reduction is very different

* add PR number to changelog

* filled in with memory number on 1.9

* addressed comments

* update comments

* fix for 1.6

* add a todo
Co-authored-by: Min Xu <min.xu@acm.org>

* [chore] 0.3.6 release

* try redo the caches
Co-authored-by: Min Xu <min.xu@acm.org>

* [chore] 0.3.5 release

* address comment
Co-authored-by: Min Xu <min.xu@acm.org>

- releasing 0.3.3
- I need it in vissl for the auto_wrap_bn change

* [feat] FSDP: add auto_wrap_bn

- add an utility function to handle wrapping of BN

* changelog

* [feature] FSDP: enable pytorch SyncBN

- not fully validated yet but at least not asserting
- this enables VISSL to move forward with its next PR

* add the test file

* changelog and lint

* addressed comment

* FSDP: multi-pass autograd graph and mixed precision

- added BACKWARD_PRE/POST checking
- better assert_state
- fixed issue of backward hook misfiring

* fix

* cleanup

* Update fairscale/nn/data_parallel/fully_sharded_data_parallel.py
Co-authored-by: Myle Ott <myleott@fb.com>
Co-authored-by: Myle Ott <myleott@fb.com>

* Adding a hard sync barrier before the broadcast, mostly useful for Gloo actually, NCCL is synced behind the scene
* adding a proper unit test
* adding a unit test for https://github.com/facebookresearch/fairscale/pull/510

* [chore] 0.3.1 release

- mainly because vissl needs the new version
- added a doc on release steps

* Update CHANGELOG.md
Co-authored-by: anj-s <32556631+anj-s@users.noreply.github.com>

* review comments
Co-authored-by: anj-s <32556631+anj-s@users.noreply.github.com>

* bring back a fix from FSDP, may help a few existing users

* v0.3.0 it is, celebration time

* POC, testing against the DDP comm hook when available
* docs, adding a reference to DDP's compress hook
* updating changelog, prep for v0.1.8 release

* [bug]: not all CUDA memory is freed when model is deleted

* fixed memory leak

- without this, peak memory will be high when more than one model
  is trained (i.e. first model leave staff around pushing up the
  peak memory when the second model runs)

* addressed comments

* fix

* changelog

* adding an assert + corresponding unit test
* updated changelog
* adjusting the adascale tests

Co-authored-by: Min Xu <24926999+min-xu-ai@users.noreply.github.com>

* [fix] ShardedDDP train/eval modes
* Update CHANGELOG.md

* initial implementation, with unit test and assert
* added changelog and better debug string

* Better unit testing
* Make it possible to refresh the DDP assumptions when the model has changed. Make it optional so that you save some time
* Enabling accumulation tests

* v0.1.6

* adding a test to prove the inter operability with upstream pytorch
* updating the changelog
* eager state pruning
* pytorch 1.5 compat

* trying to fix the missing files in the pip package (not in this diff)
* adding a long description, more pypi friendly

release pip package to follow suit

* [feat] sync adascale from internal repo

- tbd

testing: tbd

* Update argument document of __init__

* update documentation around set_num_gradients_to_accumulate

* added checking code for proper API calling places

* rename internal APIs to make them internal

* updated changelog

* added support for add_param_group and its unit test

* added unit test for set_num_gradients_to_accumulate

* added debias_ewma unit test

* fixed test_set_num_gradients_to_accumulate (need zero_grad() call)

* added missing zero_grad() to test_lr_scheduler

* fixed test_add_param_group with respect to optim.zero_grad()

* added test_gradient_value

* added test_scale_not_equal_default for scale != world_size * grad_accum

* added test_unhook()

* removed print statements

* fixed a typo

* addressed Ben's comment

* removing a dead call since ShardedDDP, small speedup
* unrelated, but filling in the changelog
* another nit

* Update changelog

missed this item from previous AdaScale commit.

* More change log

* Addressed review comments

* added AdaScale to README

* [adascale] added gradient accumulation

- added gradient accumulation
- tested with cifar full trainings with different value of accumulation
and verified the full accuracy is obtained
- also removed the patch optimize flag until we need it

* [adascale] adding pytest

- added basic and ddp tests and grad_accum
- closes #195

* added changelog

* added ddp grad_accum test

* moved ddp and non-ddp tests into separate files

* added checkpoint test

* more doc

* addressed Mike's comments

Fixes #190