[Traceable FSDP2] Ignore FSDP2 forward hook side-effects in AC; Support FSDP2 + AC #134997
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
… graph intermediate [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/134997
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (2 Unrelated Failures)As of commit 7652b37 with merge base 8d68a02 ( FLAKY - The following jobs failed but were likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
ciflow/inductor
module: inductor
oncall: distributed
… usage with graph intermediate" cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]
yf225
changed the title
… usage with graph intermediate; FSDP2 + AC support" cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]
… usage with graph intermediate; FSDP2 + AC support" cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
… usage with graph intermediate; FSDP2 + AC support" cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
… usage with graph intermediate; FSDP2 + AC support" cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
… usage with graph intermediate; FSDP2 + AC support" cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
… usage with graph intermediate; FSDP2 + AC support" cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
… usage with graph intermediate; FSDP2 + AC support" Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
… usage with graph intermediate; FSDP2 + AC support" Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
… usage with graph intermediate; FSDP2 + AC support" Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
…n AC; Support FSDP2 + AC" > Ignore FSDP2 forward hook side-effects in AC Under AC, FSDP2 does not rely on forward hook to all-gather weights to do recomputation, instead it relies on pre-backward hook to do this job: https://github.com/pytorch/pytorch/blob/451eaf0ff247090ca5a9648fd1e17c3c011737e1/torch/distributed/_composable/fsdp/_fsdp_state.py#L219-L220 So when we use `speculate_subgraph` to trace the utils.checkpoint AC region, we don't actually need to worry about FSDP2 forward hook's side effects and can safely ignore it, because we are not and we don't expect to re-run the FSDP2 forward hook during backward recomputation. ---- Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
…n AC; Support FSDP2 + AC" > Ignore FSDP2 forward hook side-effects in AC Under AC, FSDP2 does not rely on forward hook to all-gather weights to do recomputation, instead it relies on pre-backward hook to do this job: https://github.com/pytorch/pytorch/blob/451eaf0ff247090ca5a9648fd1e17c3c011737e1/torch/distributed/_composable/fsdp/_fsdp_state.py#L219-L220 So when we use `speculate_subgraph` to trace the utils.checkpoint AC region, we don't actually need to worry about FSDP2 forward hook's side effects and can safely ignore it, because we are not and we don't expect to re-run the FSDP2 forward hook during backward recomputation. ---- Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
…n AC; Support FSDP2 + AC" > Ignore FSDP2 forward hook side-effects in AC Under AC, FSDP2 does not rely on forward hook to all-gather weights to do recomputation, instead it relies on pre-backward hook to do this job: https://github.com/pytorch/pytorch/blob/451eaf0ff247090ca5a9648fd1e17c3c011737e1/torch/distributed/_composable/fsdp/_fsdp_state.py#L219-L220 So when we use `speculate_subgraph` to trace the utils.checkpoint AC region, we don't actually need to worry about FSDP2 forward hook's side effects and can safely ignore it, because we are not and we don't expect to re-run the FSDP2 forward hook during backward recomputation. ---- Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
…n AC; Support FSDP2 + AC" > Ignore FSDP2 forward hook side-effects in AC Under AC, FSDP2 does not rely on forward hook to all-gather weights to do recomputation, instead it relies on pre-backward hook to do this job: https://github.com/pytorch/pytorch/blob/451eaf0ff247090ca5a9648fd1e17c3c011737e1/torch/distributed/_composable/fsdp/_fsdp_state.py#L219-L220 So when we use `speculate_subgraph` to trace the utils.checkpoint AC region, we don't actually need to worry about FSDP2 forward hook's side effects and can safely ignore it, because we are not and we don't expect to re-run the FSDP2 forward hook during backward recomputation. ---- Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
…n AC; Support FSDP2 + AC" > Ignore FSDP2 forward hook side-effects in AC Under AC, FSDP2 does not rely on forward hook to all-gather weights to do recomputation, instead it relies on pre-backward hook to do this job: https://github.com/pytorch/pytorch/blob/451eaf0ff247090ca5a9648fd1e17c3c011737e1/torch/distributed/_composable/fsdp/_fsdp_state.py#L219-L220 So when we use `speculate_subgraph` to trace the utils.checkpoint AC region, we don't actually need to worry about FSDP2 forward hook's side effects and can safely ignore it, because we are not and we don't expect to re-run the FSDP2 forward hook during backward recomputation. ---- Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
|
The merge job was canceled or timed out. This most often happen if two merge requests were issued for the same PR, or if merge job was waiting for more than 6 hours for tests to finish. In later case, please do not hesitate to reissue the merge command |
…n AC; Support FSDP2 + AC" > Ignore FSDP2 forward hook side-effects in AC Under AC, FSDP2 does not rely on forward hook to all-gather weights to do recomputation, instead it relies on pre-backward hook to do this job: https://github.com/pytorch/pytorch/blob/451eaf0ff247090ca5a9648fd1e17c3c011737e1/torch/distributed/_composable/fsdp/_fsdp_state.py#L219-L220 So when we use `speculate_subgraph` to trace the utils.checkpoint AC region, we don't actually need to worry about FSDP2 forward hook's side effects and can safely ignore it, because we are not and we don't expect to re-run the FSDP2 forward hook during backward recomputation. ---- Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
…n AC; Support FSDP2 + AC" > Ignore FSDP2 forward hook side-effects in AC Under AC, FSDP2 does not rely on forward hook to all-gather weights to do recomputation, instead it relies on pre-backward hook to do this job: https://github.com/pytorch/pytorch/blob/451eaf0ff247090ca5a9648fd1e17c3c011737e1/torch/distributed/_composable/fsdp/_fsdp_state.py#L219-L220 So when we use `speculate_subgraph` to trace the utils.checkpoint AC region, we don't actually need to worry about FSDP2 forward hook's side effects and can safely ignore it, because we are not and we don't expect to re-run the FSDP2 forward hook during backward recomputation. ---- Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` cc XilunWu H-Huang awgu kwen2501 wanchaol fegin fduwjj wz337 wconstab d4l3k c-p-i-o voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang rec [ghstack-poisoned]
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Chao1Han
pushed a commit
to Chao1Han/pytorch
that referenced
this pull request
Sep 20, 2024
…rt FSDP2 + AC (pytorch#134997) > Ignore FSDP2 forward hook side-effects in AC Under AC, FSDP2 does not rely on forward hook to all-gather weights to do recomputation, instead it relies on pre-backward hook to do this job: https://github.com/pytorch/pytorch/blob/451eaf0ff247090ca5a9648fd1e17c3c011737e1/torch/distributed/_composable/fsdp/_fsdp_state.py#L219-L220 So when we use `speculate_subgraph` to trace the utils.checkpoint AC region, we don't actually need to worry about FSDP2 forward hook's side effects and can safely ignore it, because we are not and we don't expect to re-run the FSDP2 forward hook during backward recomputation. ---- Test commands: - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductor` - `pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductor` Pull Request resolved: pytorch#134997 Approved by: https://github.com/zou3519 ghstack dependencies: pytorch#135727
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Under AC, FSDP2 does not rely on forward hook to all-gather weights to do recomputation, instead it relies on pre-backward hook to do this job:
pytorch/torch/distributed/_composable/fsdp/_fsdp_state.py
Lines 219 to 220 in 451eaf0
So when we use
speculate_subgraphto trace the utils.checkpoint AC region, we don't actually need to worry about FSDP2 forward hook's side effects and can safely ignore it, because we are not and we don't expect to re-run the FSDP2 forward hook during backward recomputation.Test commands:
pytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_nested_fully_shard_backend_inductorpytest -rA test/distributed/_composable/fsdp/test_fully_shard_compile.py::TestFullyShardCompile::test_transformer_backend_inductorStack from ghstack (oldest at bottom):
cc @XilunWu @H-Huang @awgu @kwen2501 @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @c-p-i-o @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang @rec