Ensure gradient clear out pending AsyncCollectiveTensor in FSDP Extension (pytorch/pytorch#116122)
Fix processing unflatten tensor on compute stream in FSDP Extension (pytorch/pytorch#116559)
Fix FSDP AssertionError on tensor subclass when setting sync_module_states=True (pytorch/pytorch#117336)
Fix DCP state_dict cannot correctly find FQN when the leaf module is wrapped by FSDP (pytorch/pytorch#115592)
Fix OOM when when returning a AsyncCollectiveTensor by forcing _gather_state_dict() to be synchronous with respect to the mian stream. (pytorch/pytorch#118197) (pytorch/pytorch#119716)
Fix Windows runtime torch.distributed.DistNetworkError: [WinError 32] The process cannot access the file because it is being used by another process (pytorch/pytorch#118860)
Fix backward compatibility issues with Llama and Gemma:
We mostly made sure that performances are not affected by the new change of paradigm with ROPE. Fixed the ROPE computation (should always be in float32) and the causal_mask dtype was set to bool to take less RAM.
YOLOS had a regression, and Llama / T5Tokenizer had a warning popping for random reasons
FIX [Gemma] Fix bad rebase with transformers main (#29170)
Gemma is a new opensource Language Model series from Google AI that comes with a 2B and 7B variant. The release comes with the pre-trained and instruction fine-tuned versions and you can use them via AutoModelForCausalLM, GemmaForCausalLM or pipeline interface!
#11895: Fix collection on Windows where initial paths contain the short version of a path (for example c:\PROGRA~1\tests).
#11953: Fix an IndexError crash raising from getstatementrange_ast.
#12021: Reverted a fix to [--maxfail]{.title-ref} handling in pytest 8.0.0 because it caused a regression in pytest-xdist whereby session fixture teardowns may get executed multiple times when the max-fails is reached.
8.0.1
pytest 8.0.1 (2024-02-16)
Bug Fixes
#11875: Correctly handle errors from getpass.getuser{.interpreted-text role="func"} in Python 3.13.
#11879: Fix an edge case where ExceptionInfo._stringify_exception could crash pytest.raises{.interpreted-text role="func"}.
#11906: Fix regression with pytest.warns{.interpreted-text role="func"} using custom warning subclasses which have more than one parameter in their [__init__]{.title-ref}.
#11907: Fix a regression in pytest 8.0.0 whereby calling pytest.skip{.interpreted-text role="func"} and similar control-flow exceptions within a pytest.warns(){.interpreted-text role="func"} block would get suppressed instead of propagating.
#11929: Fix a regression in pytest 8.0.0 whereby autouse fixtures defined in a module get ignored by the doctests in the module.
#11937: Fix a regression in pytest 8.0.0 whereby items would be collected in reverse order in some circumstances.
Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.
Dependabot commands and options
You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it
- `@dependabot merge` will merge this PR after your CI passes on it
- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it
- `@dependabot cancel merge` will cancel a previously requested merge and block automerging
- `@dependabot reopen` will reopen this PR if it is closed
- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually
- `@dependabot show ignore conditions` will show all of the ignore conditions of the specified dependency
- `@dependabot ignore major version` will close this group update PR and stop Dependabot creating any more for the specific dependency's major version (unless you unignore this specific dependency's major version or upgrade to it yourself)
- `@dependabot ignore minor version` will close this group update PR and stop Dependabot creating any more for the specific dependency's minor version (unless you unignore this specific dependency's minor version or upgrade to it yourself)
- `@dependabot ignore ` will close this group update PR and stop Dependabot creating any more for the specific dependency (unless you unignore this specific dependency or upgrade to it yourself)
- `@dependabot unignore ` will remove all of the ignore conditions of the specified dependency
- `@dependabot unignore ` will remove the ignore condition of the specified dependency and ignore conditions
Bumps the python-packages group with 5 updates:
3.1.41
3.1.42
2.2.0
2.2.1
1.10.0
1.13.3
4.37.2
4.38.2
8.0.0
8.0.2
Updates
gitpython
from 3.1.41 to 3.1.42Release notes
Sourced from gitpython's releases.
Commits
1f37b48
prepare the next release9caf3ae
Merge pull request #1825 from EliahKagan/tree-test2613421
Merge pull request #1823 from marcm-ml/masterb780a8c
Tweak@with_rw_directory
and go back to using it0114a99
Use more ligtweight approach to guarantee deletion90cf4d7
Fix new PermissionError in Windows with Python 3.7dd42e38
Keep temp files out of project dir and improve cleanup2671167
Remove deprecated section in README.md7ba3fd2
Bump Vampire/setup-wsl from 2.0.2 to 3.0.0e75ea98
Bump pre-commit/action from 3.0.0 to 3.0.1Updates
torch
from 2.2.0 to 2.2.1Release notes
Sourced from torch's releases.
Commits
6c8c5ad
[RelEng] DefineBUILD_BUNDLE_PTXAS
(#119750) (#119988)f00f0ab
fix compile DTensor.from_local in trace_rule_look up (#119659) (#119941)077791b
Revert "Update state_dict.py to propagate cpu offload (#117453)" (#119995)3eaaeeb
Update state_dict.py to propagate cpu offload (#117453) (#119916)0aa3fd3
HSDP + TP integration bug fixes (#119819)eef51a6
[Inductor] Skip triton templates for mixedmm on SM70- (#118591) (#119894)940358f
[dtensor] fix dtensor _to_copy op for mix precision (#116426) (#119687)24e4751
[state_dict] Calls wait() for the DTensor to_local() result (#118197) (#119692)dcaeed3
[DCP][state_dict] Fix the issue that get_state_dict/set_state_dict ig… (#119807)4f882a5
Properly preserve SymInt input invariant when splitting graphs (#117406) (#11...Updates
openai
from 1.10.0 to 1.13.3Release notes
Sourced from openai's releases.
... (truncated)
Changelog
Sourced from openai's changelog.
... (truncated)
Commits
e41abf7
release: 1.13.3cbcb978
chore(client): use anyio.sleep instead of asyncio.sleep (#1198)9375d2c
docs(examples): add pyaudio streaming example (#1194)9179a03
chore(internal): bump pyright (#1193)6d25c6b
feat(api): add wav and pcm to response_format (#1189)af02745
docs: add note in azure_deployment docstring (#1188)f0b1823
chore(types): extract run status to a named type (#1178)a7115b5
release: 1.13.263222f6
fix(ci): revert "move github release logic to github app" (#1170)f504384
release: 1.13.1Updates
transformers
from 4.37.2 to 4.38.2Release notes
Sourced from transformers's releases.
... (truncated)
Commits
092f1fd
Release 4.38.2bf5163f
fix merge conflicts between llama and gemma6c45f0f
Usetorch.bool
instead oftorch.int64
for non-persistant causal mask buff...bfefb8e
Patch YOLOS and others (#29353)20164cc
RoPE loses precision for Llama / Gemma + Gemma logits.float() (#29285)d5ec194
[Llama ROPE
] Fix torch export but also slow downs in forward (#29198)bf99e86
[T5 and Llama Tokenizer
] remove warning (#29346)6d02350
Improve _update_causal_mask performance (#29210)4f8689e
FIX [Gemma
] Fix bad rebase with transformers main (#29170)a085774
Release: v4.38.1Updates
pytest
from 8.0.0 to 8.0.2Release notes
Sourced from pytest's releases.
Commits
31afeeb
Prepare release version 8.0.21b00a2f
Merge pull request #12025 from pytest-dev/backport-12022-to-8.0.xff2f66d
[8.0.x] Revert "Fix teardown error reporting when--maxfail=1
(#11721)"8a8eed6
[8.0.x] Fix collection of short paths on Windows (#12024)74346f0
[8.0.x] Allow Sphinx 7.x (#12005)b7657b4
[8.0.x] Disallow Sphinx 6 and 7 (#12001)feb7c5e
Merge pull request #11999 from pytest-dev/backport-11996-to-8.0.x0909655
[8.0.x] code: fixIndexError
crash ingetstatementrange_ast
68524d4
Merge pull request #11993 from pytest-dev/release-8.0.1d7d320a
Prepare release version 8.0.1Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting
@dependabot rebase
.Dependabot commands and options
You can trigger Dependabot actions by commenting on this PR: - `@dependabot rebase` will rebase this PR - `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it - `@dependabot merge` will merge this PR after your CI passes on it - `@dependabot squash and merge` will squash and merge this PR after your CI passes on it - `@dependabot cancel merge` will cancel a previously requested merge and block automerging - `@dependabot reopen` will reopen this PR if it is closed - `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually - `@dependabot show