| .. |
|
_composable
|
Back out "[compiled autograd] tls access helpers (#138061)" and Back out "[compiled autograd] Compiled autograd configs in TLS (#137821)" (#139086)
|
2024-10-28 23:37:05 +00:00 |
|
_shard
|
|
|
|
_tensor
|
Fix unbind_copy and add its decomposition (#134319)
|
2024-10-23 19:13:44 +00:00 |
|
_tools
|
Upgrade distributed test to g4dn instances (T4 GPUs) (#137161)
|
2024-10-20 23:48:54 +00:00 |
|
algorithms
|
|
|
|
bin
|
|
|
|
checkpoint
|
[Distributed][Test] Fix todo in distributed test files (#136836)
|
2024-10-16 01:15:12 +00:00 |
|
elastic
|
|
|
|
flight_recorder
|
|
|
|
fsdp
|
[Distributed][Test] Fix todo in distributed test files (#136836)
|
2024-10-16 01:15:12 +00:00 |
|
launcher
|
|
|
|
nn/jit
|
|
|
|
optim
|
|
|
|
pipelining
|
[Pipelining] small comments and variable renames (#138735)
|
2024-10-25 12:06:54 +00:00 |
|
rpc
|
|
|
|
tensor/parallel
|
[Inductor] improve the stride preservation logic of user-visible outputs (#136732)
|
2024-10-26 18:49:14 +00:00 |
|
argparse_util_test.py
|
|
|
|
test_c10d_common.py
|
|
|
|
test_c10d_functional_native.py
|
Revert "[c10d][Partial-Graph Overlap] Support calling .wait_tensor() on output tensor of eager async_op=True collective if under allow_inflight_collective_as_graph_input_ctx() context manager (#137763)"
|
2024-10-28 20:13:46 +00:00 |
|
test_c10d_gloo.py
|
|
|
|
test_c10d_logger.py
|
|
|
|
test_c10d_nccl.py
|
Revert "[PGNCCL] Make sure we do not use split for P2P comm creation (#139013)"
|
2024-10-28 21:30:28 +00:00 |
|
test_c10d_object_collectives.py
|
[c10d][CI] Improve world size setting in some tests (#138846)
|
2024-10-25 23:02:17 +00:00 |
|
test_c10d_ops_nccl.py
|
[c10d][CI] Improve world size setting in some tests (#138846)
|
2024-10-25 23:02:17 +00:00 |
|
test_c10d_pypg.py
|
|
|
|
test_c10d_spawn.py
|
|
|
|
test_c10d_spawn_gloo.py
|
|
|
|
test_c10d_spawn_nccl.py
|
|
|
|
test_c10d_spawn_ucc.py
|
|
|
|
test_c10d_ucc.py
|
|
|
|
test_collective_utils.py
|
|
|
|
test_compute_comm_reordering.py
|
[CI] Add Compiled DDP / Compiled FSDP2 / compute-comm reordering tests to test_inductor_distributed (#138178)
|
2024-10-20 19:38:18 +00:00 |
|
test_control_collectives.py
|
|
|
|
test_data_parallel.py
|
|
|
|
test_device_mesh.py
|
[DeviceMesh] Use split_group to create sub_groups for nccl backend if the default pg is eagerly initialized (#138129)
|
2024-10-22 00:00:05 +00:00 |
|
test_distributed_spawn.py
|
|
|
|
test_dynamo_distributed.py
|
Made DDPOptimizer work with HOPs (#138787)
|
2024-10-25 18:59:01 +00:00 |
|
test_fake_pg.py
|
|
|
|
test_functional_api.py
|
|
|
|
test_inductor_collectives.py
|
Revert "[c10d][Partial-Graph Overlap] Support calling .wait_tensor() on output tensor of eager async_op=True collective if under allow_inflight_collective_as_graph_input_ctx() context manager (#137763)"
|
2024-10-28 20:13:46 +00:00 |
|
test_launcher.py
|
|
|
|
test_multi_threaded_pg.py
|
|
|
|
test_nccl.py
|
[Pytorch][ATEN] Enable FP8 NCCL in Pytorch ATEN (#138776)
|
2024-10-25 21:56:47 +00:00 |
|
test_pg_wrapper.py
|
|
|
|
test_store.py
|
[TCPStore] Throw value error if passing world_size=0 to TCPStore (#137792)
|
2024-10-11 23:42:57 +00:00 |
|
test_symmetric_memory.py
|
[fused_all_gather_scaled_matmul] support rowwise scaling (#137805)
|
2024-10-15 21:35:14 +00:00 |
