[TPU] TorchTPU backend integration - eager / torch.compile / tp - #14039

Open

JingyaHuang wants to merge 40 commits into

huggingface:mainfrom

JingyaHuang:add-torchtpu-support

Commits on May 29, 2026

feat: add torchtpu
JingyaHuang
committed
38007b4

Commits on Jun 2, 2026

feat:draft TorchTPU support
JingyaHuang
committed
8ed15f9

Commits on Jun 9, 2026

fix: wan overflow issue + compile mode error on sdxl
JingyaHuang
committed
e343c0a

Commits on Jun 25, 2026

doc: enhance with TorchTPU doc
JingyaHuang
committed
84b4049
doc: enhance with TorchTPU doc
JingyaHuang
committed
bb3ec1e
Merge branch 'main' into add-torchtpu-support
JingyaHuang
authored
339be41

Commits on Jul 17, 2026

Commits on Jul 27, 2026

Merge branch 'main' into add-torchtpu-support
JingyaHuang
authored
05d56d5

Commits on Sep 2, 2026

Merge branch 'main' into add-torchtpu-support
JingyaHuang
authored
b7a8c0b

Commits on Sep 3, 2026

Merge branch 'huggingface:main' into add-torchtpu-support
JingyaHuang
authored
b0c5595

Commits on Sep 7, 2026

Merge branch 'main' into add-torchtpu-support
JingyaHuang
authored
7c3df6c
feat: add TP for TPU

JingyaHuang
authored and
JingyaHuang
committed
7c1e379
style: fix import sorting in TPU test scripts

authored and
JingyaHuang
committed
f944c48
style: ruff format TPU test scripts

authored and
JingyaHuang
committed
82d20ab
fix: style
JingyaHuang
committed
f6c17d2
fix: test for native 4 devices
JingyaHuang
committed
e8fe48c

Commits on Sep 8, 2026

fix: propagate TPU device fixes to Flux/Flux2/Wan-family copies; register tpu.md; drop redundant execution_device check

JingyaHuang
and
claude
committed
9b62254
tests: cleanup
JingyaHuang
committed
744fa95
fix: fix compile mode
JingyaHuang
committed
bfc7d17

Commits on Sep 9, 2026

Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
10d4ad1
Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
a999a3d
Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
cb2dcf3
Merge branch 'main' into add-torchtpu-support
JingyaHuang
authored
9f11d1e
Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
c48d41c
Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
aa70a45
Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
6cb37d2
Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
3e06f71
Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
09d5ce2
Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
7b46e08
Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
e82e795
Update docs/source/en/optimization/tpu.md

JingyaHuang
and
stevhliu
authored
d2bf529
test: remove flux2 e2e test
JingyaHuang
committed
22d595a
doc: apply suggestions
JingyaHuang
committed
042a88a
doc: apply suggestions
JingyaHuang
committed
c6a4c80
Merge branch 'add-torchtpu-support' of github.com:JingyaHuang/diffusers into add-torchtpu-support
JingyaHuang
committed
c82fce7

Commits on Sep 10, 2026

review: remove monkey patch
JingyaHuang
committed
cb314bd
review: revert neuron-specific changes in the tests
JingyaHuang
committed
bf8934f