Bump the python group with 2 updates

[dependabot]

Bumps the python group with 2 updates: torch and torch-xla.

Updates torch from 2.6.0 to 2.7.0

Release notes

Sourced from torch's releases.

PyTorch 2.7.0 Release Notes

• Highlights
• Tracked Regressions
• Backwards Incompatible Changes
• Deprecations
• New Features
• Improvements
• Bug fixes
• Performance
• Documentation
• Developers

Highlights

... (truncated)

Commits

• 1341794 Gracefully handle optree less than minimum version, part 2 (#151323)
• 0739127 Gracefully handle optree less than minimum version (#150977)
• 0c236f3 Update triton wheel build, setuptools pin (#150953)
• c7ff78d Fix inplacing with multiple, fused uses (#150892)
• 894909a Revert "[CUDA] Only use vec128 if CUDA version is newer than 12.8" (#150855)
• ef2b139 [Manylinux 2.28] Correct Linux aarch64 cuda binaries wheel name (#150820)
• 3f236f1 [CUDA] Only use vec128 if CUDA version is newer than 12.8 (#150818)
• 35f1e76 Reland of "[ROCm] change preferred blas lib defaults (#150249)"" (#150707)
• a6321d6 Revert "Dont exclude constant_pad_nd in prologue fusion" (#150699)
• 1cc51c6 [CUDA][avgpool2d] Fix backward launch bounds again for sm100, sm120 (#150...
• Additional commits viewable in compare view

Updates torch-xla from 2.6.0 to 2.7.0

Release notes

Sourced from torch-xla's releases.

PyTorch/XLA 2.7 release

Highlights

• Easier training on Cloud TPUs with TorchPrime
• A new Pallas-based kernel for ragged paged attention, enabling further optimizations on vLLM TPU (#8791)
• Usability improvements
• Experimental JAX interoperability with JAX operations (#8781, #8789, #8830, #8878)
• re-enabled GPU CI build [#8593]

Stable Features

• Operator Lowering
  • Lower as_strided_copy to use fast path with slice (#8374)
  • Lower _conj_copy. (#8686)
• Support splitting physical axis in SPMD mesh (#8698)
• Support of placeholder tensor (#8785).
• Dynamo/AOTAutograd traceable flash attention(#8654)
• C++11 ABI build is the default

Experimental Features

• Gated Recurrent Unit (GRU) implemented with scan (#8777)
• Introduce apply_xla_patch_to_nn_linear to improve einsum performance (#8793)
• Enable default buffer donation for step barriers (#8721, #8982)

Usability

• Better profiling control: the start and the end of the profiling session can be controlled by the new profiler API (#8743)
• API to query number of cached compilation graphs (#8822)
• Enhancement on host-to-device transfer (#8849)

Bug fixes

• fix a bug in tensor.flatten (#8680)
• cummax: fix 0-sized dimension reduction. (#8653)
• Fix dk/dv autograd error on TPU flash attention (#8685)
• Fix a bug in flash attention where kv_seq_len should divide block_k_major. (#8671)
• [scan] Make sure inputs into fn are not device_data IR nodes(#8769)

Libtpu stable version

• Pin 2.7 release to stable libtpu version '0.0.11.1'

Deprecations

• Deprecate torch.export and instead, use torchax to export graph to StableHLO for full dynamism support
• Remove torch_xla.core.xla_model.xrt_world_size, replace with torch_xla.runtime.world_size
• Remove torch_xla.core.xla_model.get_ordinal, replace with torch_xla.runtime.global_ordinal
• Remove torch_xla.core.xla_model.parse_xla_device, replace with _utils.parse_xla_device
• Remove torch_xla.experimental.compile, replace with torch_xla.compile

Commits

• d5871fc support manylinux build (#9035)
• 839ac4c Fix 3.9 Python syntax (#9018)
• 5d3c70e Remove unnecessary registries for 2.7 install instructions (#9012)
• ab93be8 Remove config warning log from GradAcc API (#9007)
• ee35d13 Default explicit donation for step barriers (#9005)
• 124075e Update libtpu release path (#9009)
• 9482006 Fix the source param sharding for GradAcc API (#8999) (#9000)
• 5d3ecd8 Set scoped vmem for paged attention (#8988) (#8994)
• 4a482db cherry pick a37a5852808861a6b274cfbd1a98653fd1bc6a9a (#8995)
• 65fc779 Avoid re-computing computation hashes (#8976) (#8977)
• Additional commits viewable in compare view

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 <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore <dependency name> 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 <dependency name> 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 <dependency name> 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 <dependency name> will remove all of the ignore conditions of the specified dependency
• @dependabot unignore <dependency name> <ignore condition> will remove the ignore condition of the specified dependency and ignore conditions