Sherlock Huang 4ac4a7351e Shortcut redistribution when num_shards == 1 (#163742) ...

Redistribution doesn't need collectives when num_shards == 1 on a mesh dimension.
Only placement update is needed, local_tensor remains unchanged.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/163742
Approved by: https://github.com/tianyu-l

Co-authored-by: tianyu-l <150487191+tianyu-l@users.noreply.github.com>

2025-09-24 23:49:08 +00:00

..

_composable

Add early_stop kwarg to torch.utils.checkpoint (#160781)

2025-08-26 22:32:35 +00:00

_pycute

[CuTe] Change the logic of pycute manipulation ops like coalesce, complement from co-lex to lex (#162690)

2025-09-16 19:53:45 +00:00

_shard

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

_sharded_tensor

[BE][PYFMT] migrate PYFMT for torch.{distributed,distributions} to ruff format (#144547)

2025-02-28 07:35:56 +00:00

_sharding_spec

[BE][Easy] enable UFMT for torch/distributed/ (#128870)

2024-06-22 18:53:28 +00:00

_symmetric_memory

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

_tensor

[BE][PYFMT] migrate PYFMT for torch.{distributed,distributions} to ruff format (#144547)

2025-02-28 07:35:56 +00:00

_tools

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

algorithms

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

autograd

[remove untyped defs] batch 1 (#157011)

2025-06-30 23:54:40 +00:00

benchmarks

[BE][CI] bump ruff to 0.8.4 (#143753)

2024-12-24 12:24:10 +00:00

checkpoint

[DCP] DTensor slice dequantization with proper block alignment (#163532)

2025-09-23 16:48:16 +00:00

elastic

[BE] Delete all pre py-3.10 checks (#163653)

2025-09-23 23:22:53 +00:00

examples

Support XPU in memory tracker (#150703)

2025-06-12 21:33:52 +00:00

fsdp

[FSDP2] idempotent reset_sharded_param: no-op if _local_tensor is already padded (#163130)

2025-09-18 09:20:37 +00:00

launcher

154849 Add support to handle IGUSR1 and SIGUSR2 in multiprocessing (#160690)

2025-09-09 22:23:06 +00:00

nn

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

optim

[BE][5/16] fix typos in torch/ (torch/distributed/) (#156315)

2025-06-23 02:57:28 +00:00

pipelining

Inspect schedule IR comms (#162996)

2025-09-16 16:59:06 +00:00

rpc

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

tensor

Shortcut redistribution when num_shards == 1 (#163742)

2025-09-24 23:49:08 +00:00

__init__.py

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

_C_stubs.py

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

_checkpointable.py

[BE]: Backport runtime_checkable perf improvements/behavior from 3.12 (#155130)

2025-06-06 13:28:05 +00:00

_composable_state.py

[FSDP2] Make module-to-state mapping use weakrefs (#139650)

2024-11-05 02:16:52 +00:00

_dist2.py

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

_distributed_c10d.py

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

_functional_collectives_impl.py

PEP585 update - torch/distributed (#145164)

2025-01-21 04:23:29 +00:00

_functional_collectives.py

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

_mesh_layout.py

[DeviceMesh] Make CuTe layout as mesh layout to be ready for using in DeviceMesh (#162414)

2025-09-15 17:04:41 +00:00

_serialization.py

[BE][5/16] fix typos in torch/ (torch/distributed/) (#156315)

2025-06-23 02:57:28 +00:00

_state_dict_utils.py

fix-unpin-memory-tensor-param (#160992)

2025-08-26 21:55:25 +00:00

argparse_util.py

Flip default value for mypy disallow_untyped_defs [5/11] (#127842)

2024-06-08 18:49:18 +00:00

c10d_logger.py

PEP585 update - torch/distributed (#145164)

2025-01-21 04:23:29 +00:00

collective_utils.py

[C10D] add _summarize_ranks util (#160284)

2025-08-28 00:17:53 +00:00

constants.py

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

CONTRIBUTING.md

fix torch/distributed contributing doc (#158934)

2025-07-28 17:01:05 +00:00

device_mesh.py

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

distributed_c10d.py

[RELAND] Always build USE_DISTRIBUTED (#160449) and Make distributed modules importable even when backend not built (#159889) (#162594)

2025-09-22 21:12:18 +00:00

launch.py

[BE][PYFMT] migrate PYFMT for torch.{distributed,distributions} to ruff format (#144547)

2025-02-28 07:35:56 +00:00

logging_handlers.py

PEP585 update - torch/distributed (#145164)

2025-01-21 04:23:29 +00:00

remote_device.py

[BE][Easy] fix ruff rule needless-bool (SIM103) (#130206)

2024-07-14 08:17:52 +00:00

rendezvous.py

[BE][5/16] fix typos in torch/ (torch/distributed/) (#156315)

2025-06-23 02:57:28 +00:00

run.py

Support XPU in --nproc-per-node option to torchrun (#159474)

2025-09-12 08:32:04 +00:00

utils.py

Refactor to use torch.accelerator.device_index instead of torch.cuda.device for generic device context manager (#148880)

2025-04-25 09:45:25 +00:00