[v2.3.1] Release Tracker about pytorch HOT 29 CLOSED

⚠️ #126567 must be merged before

Link to landed trunk PR:

• #125337

Link to release branch PR:

• #127219

Criteria Category:

• Low risk critical fix for checkpointing error introduced in 2.3. Fixes #122792

@atalman wrote: Hi @antoinebrl the PR: #126567 was merged, please correct posted cherry-pick
@antoinebrl: Thanks @atalman! I isolated the cherry picked commit now that the previous refactoring is merged.
@huydhn merged

from pytorch.

Please note, we are in the phase:
Phase 2 (after 5/27): Perform extended integration/stability/performance testing based on Release Candidate builds.
We do not accept any new cherry-picks

from pytorch.

Link to main PR (if applicable):

• pytorch/builder#1803

Link to 2.3.1 release branch PR:

• pytorch/builder#1805

Criteria Category:
Fixes performance regression issues pytorch/builder#1774 and #124922

@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):

• #125265

Link to release branch PR:

• #125443

Criteria Category:

• Documentation improvements

@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):

• #124634

Link to release branch PR:

• #125755

Criteria Category:

• low risk critical fix

@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):

• This is not a PR cherrypick but a revert commit: 51cf57c

Link to release branch PR:

• #125790

Criteria Category:

• Low risk critical fix

@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):

• #122616

Link to release branch PR:

• #126094

Criteria Category:

• Critical Fix - torchdata compatibility

@huydhn merged

from pytorch.

Link to landed trunk PR (if applicable):

• #125684

Link to release branch PR:

• #126095

Criteria Category:

• Critical fix for #125109

@huydhn merged

from pytorch.

Link to landed trunk PR (if applicable):

• #125662

Link to release branch PR:

• #126096

Criteria Category:

• Critical fix for silent correctness issue: #125135

@huydhn merged

from pytorch.

Link to landed trunk PR (if applicable):

• #125617

Link to release branch PR:

• #126099

Criteria Category:

• Critical fix for: #125094

@huydhn merged

from pytorch.

Link to landed trunk PR (if applicable):

• #122895

Link to release branch PR:

• #126102

Criteria Category:

• Fixes doc failure on Release branch

@huydhn merged

from pytorch.

Link to landed trunk PR (if applicable):

• #125355

Link to release branch PR:

• #126100

Criteria Category:

• Critical Fixes #124850

@huydhn merged

from pytorch.

Link to landed trunk PR (if applicable):

• #125154

Link to release branch PR:

• #126101

Criteria Category:

• Fixes Crash #124868

@huydhn I update the cherry pick manually to remove changes from #124479, let's see if that works.
@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):

• #124635

Link to release branch PR:

• #126104

Criteria Category:

• Critical fix #124335

@huydhn merged

from pytorch.

Link to landed trunk PR (if applicable):

• #124238

Link to release branch PR:

• #126235

Criteria Category:

• Critical fix #119607

@huydhn This cherry pick is complex from what I see because cherry picking it blindly won't work. It depends on #122146, that enables dynamo for 3.12. And that stack is not a small one. So, I think we need to rework the cherry pick #126107 manually if we want to go ahead with this (cc @williamwen42 @atalman @malfet @albanD)

@atalman Closed - #126107 . Merged - #126235

from pytorch.

Link to landed trunk PR (if applicable):

• #123407

Link to release branch PR:

• #123482

Criteria Category:

• Fixes new feature

@huydhn merged

from pytorch.

Link to landed trunk PR (if applicable):

• #125843

Link to release branch PR:

• #126108

Criteria Category:

• Required to fix Lint errors on release

@huydhn merged

from pytorch.

Link to landed trunk PR (if applicable):

• #124780

Link to release branch PR:

• #126147

Criteria Category:

• Fixes to regressions against the 2.3.0 release

@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):
*

Link to release branch PR:

• #126204
• #126202
• #126309
• #127290

Criteria Category:

• Release only - use triton 2.3.1 version rather then current 2.3.0. Revert temp change after triton 2.3.1 release

from pytorch.

Link to landed trunk PR (if applicable):

• #124932

Link to release branch PR:

• #126301

Criteria Category:

• Documentation

@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):

• #120756

Link to release branch PR:

• #126332

Criteria Category:

• Critical fix #119607 torch.compile refleak

@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):

• #121030

Link to release branch PR:

• #126448

Criteria Category:

• Low risk addition of hipify mappings to enable DeepSpeed transformer extensions on ROCm

@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):

• #123617

Link to release branch PR:

• #126957

Criteria Category:

• Fixes to regressions against the 2.3.0 release

@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):

• NA

Link to release branch PR:

• #126452
• #126942

Criteria Category:

• Release only changes - pin docker image for rocm CI. Temporary PR. ROCm test jobs were failing with the MIOpen error because a subtle difference crept in the MIOpen kdb files when the docker images were rebuilt. Hence pin to make CI jobs green. Will unpin once kdb issue is resolved

@atalman merged #126452

from pytorch.

Link to main PR (if applicable):

• #121544
• #127069

Link to 2.3.1 release branch PR:

• #126557
• #127130

Criteria Category:

• Low risk critical fix for checkpointing. This PR removes an additional check introduced in torch 2.3 which is actually incorrect and causes checkpointing to fail if at least 1 forward/backward pass has not been run.

@fegin Cherry-pick the test PR #127130. cc.,
@atalman merged

from pytorch.

Link to main PR (if applicable):

• #124698
• #126935

Link to 2.3.1 release branch PR:

• #126559
• #126992

Criteria Category:

• Low risk critical fix for checkpointing. When checkpointing with activation checkpointing, the names of the variables (FQNs) are changed with the tag _checkpoint_wrapped_module. This breaks checkpointing if you resume without activation checkpointing.
• @wz337 Added unit test for this cherry-pick.

@atalman merged

from pytorch.

Link to main PR:

• #125336

Link to 2.3.1 release branch PR:

• #126567

Criteria Category:

• Low risk critical fix for checkpointing. Torch 2.3 ignores _extra_state (whereas prior PyTorch versions correctly handle it), breaking checkpoint loading from prior checkpoints. This also breaks integration with Nvidia's TransformerEngine.

@atalman merged

from pytorch.

Link to landed trunk PR (if applicable):

• #126414
• #123795

Link to release branch PR:

• #127312
• #127322

Criteria Category:

• CI fix to keep inductor job and pull jobs green in release branch

from pytorch.

closing release tracker. release published

from pytorch.