mirror of
https://github.com/saymrwulf/pytorch.git
synced 2026-05-15 21:00:47 +00:00
762a05b3b3
The original `_all_gather_keys` call was for a safety check, but could be costly as things scale, and it blocks CPU. Instead, we make it clear in the documentation that the `state_dict` passed to the `load` API should have same set of keys, otherwise the API may hang. In addition, we move the check to a utility function: `utils.assert_same_keys`. User uncertain about state dict unity can optionally call this API to check. Resolves #145965 (as a workaround). Pull Request resolved: https://github.com/pytorch/pytorch/pull/145998 Approved by: https://github.com/mhorowitz, https://github.com/fegin |
||
|---|---|---|
| .. | ||
| _composable | ||
| _shard | ||
| _tools | ||
| algorithms | ||
| bin | ||
| checkpoint | ||
| elastic | ||
| flight_recorder | ||
| fsdp | ||
| launcher | ||
| nn/jit | ||
| optim | ||
| pipelining | ||
| rpc | ||
| tensor | ||
| argparse_util_test.py | ||
| test_backends.py | ||
| test_c10d_common.py | ||
| test_c10d_functional_native.py | ||
| test_c10d_gloo.py | ||
| test_c10d_logger.py | ||
| test_c10d_nccl.py | ||
| test_c10d_object_collectives.py | ||
| test_c10d_ops_nccl.py | ||
| 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_composability.py | ||
| test_compute_comm_reordering.py | ||
| test_control_collectives.py | ||
| test_data_parallel.py | ||
| test_device_mesh.py | ||
| test_distributed_spawn.py | ||
| test_dynamo_distributed.py | ||
| test_fake_pg.py | ||
| test_functional_api.py | ||
| test_inductor_collectives.py | ||
| test_launcher.py | ||
| test_multi_threaded_pg.py | ||
| test_nccl.py | ||
| test_pg_wrapper.py | ||
| test_store.py | ||
| test_symmetric_memory.py | ||