Moving DTensor to be in the public namespace, to formally add the
documentation page that includes all the public APIs. This includes:
* many path renames and path import fixes
* a dedicated doc page without too much content yet (adding in the next
PRs)
* To preserve the BC for users still using the `torch.distributed._tensor`,
I added a shim script to redirect old path calls to the new module
The BC preserving is evidented by the fact that all DTensor tests are still
working without changing the public imports. So it's safe to land the
changes
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133113
Approved by: https://github.com/XilunWu
ghstack dependencies: #133305, #133306
Previously, when we slice out a submesh from a mesh, we assign the mesh as the parent mesh of the submesh. In this case, when we have a 3D mesh topology, the parent mesh of a 1D mesh sliced out from the 3D mesh is different from the parent mesh of the same 1D mesh sliced out from the 2D submesh of the 3D mesh. For example:
```
mesh_3d = init_device_mesh("cuda", (2,2,2), ("dim0", "dim1", "dim2"))
mesh_dim0 = mesh_3d["dim0"]
mesh_2d = mesh_2d["dim0", "dim1"]
mesh_dim0_2 = mesh_2d["dim0_2"]
# This would evaluate to be True
print(_mesh_resources.get_parent_mesh(mesh_dim0) != _mesh_resources.get_parent_mesh(mesh_dim0))
```
We can always reconstruct the mesh needed from the mesh dim names, as long as two dims come from the same root. For simplicity, we do not see the necessity of building a tree structure to represent child-parent relationship. Therefore, we are replacing the parent mesh concept with a root mesh concept in `_MeshEnv` so we would have:
```
mesh_3d = init_device_mesh("cuda", (2,2,2), ("dim0", "dim1", "dim2"))
mesh_dim0 = mesh_3d["dim0"]
mesh_2d = mesh_2d["dim0", "dim1"]
mesh_dim0_2 = mesh_2d["dim0_2"]
# This would evaluate to be True
print(_mesh_resources.get_root_mesh(mesh_dim0) == _mesh_resources.get_root_mesh(mesh_dim0))
```
With this change, we will have two types of meshes in an environment.
1. `device_mesh != _mesh_resources.get_root_mesh(device_mesh)` means that the device_mesh is created by slicing.
2. `device_mesh == _mesh_resources.get_root_mesh(device_mesh)` means that the device_mesh is a root mesh not created through slicing.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132339
Approved by: https://github.com/wanchaol
ghstack dependencies: #132310, #132311
In _creating chunk_sharded_tensor, _get_remote_device_str is used. by default it uses the node cound to determine the device:instance. for hpu, need to use current device to get the deivce_instance.
Fixes #ISSUE_NUMBER
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132120
Approved by: https://github.com/awgu
This PR adds a all_gather_dtensor() method to fsdp/_fsdp_extensions.py and the actual implementation in tensor/parallel/fsdp.py. This enables FSDP to load 2D DTensor state_dict into model when calling `model.load_state_dict()`.
cc. @fegin
Pull Request resolved: https://github.com/pytorch/pytorch/pull/110925
Approved by: https://github.com/fegin
ghstack dependencies: #110831, #110846
This PR:
1) Add device_mesh kwarg to FSDP. Remove init_device_mesh() from _runtime_utils.py, as device_mesh would be passed in by user as an kwarg.
2) change use_dtensor flag for state_dict_config and optim_state_dict_config to be private. If device_mesh is used with sharded model/optim state dict, _use_dtensor flag would be set to True and model/optim state dict would return dtensor state_dict. Otherwise, _use_dtensor flag would be set to False and model/optim state dict would return sharded_tensor state_dict.
3) Update _optim_utils.py, _shard_utils.py, and _state_dict_utils.py to add support for HSDP to return 2D DTensor state_dict.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107533
Approved by: https://github.com/fegin, https://github.com/awgu, https://github.com/wanchaol
This PR adds SimpleProfiler for FSDP state_dict/load_state_dict logging purpose. SimpleProfiler use class variables to record profiling results and it does everything in the Python which can be slow. So it is only suitable for logging slow actions such as initialization and state_dict/load_state_dict.
This PR uses SimpleProfiler to log some critical/slow paths of the model and optimizer state_dict/load_state_dict.
Differential Revision: [D48774406](https://our.internmc.facebook.com/intern/diff/D48774406/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/108290
Approved by: https://github.com/wz337
This PR:
1) Add device_mesh kwarg to FSDP. Remove init_device_mesh() from _runtime_utils.py, as device_mesh would be passed in by user as an kwarg.
2) change use_dtensor flag for state_dict_config and optim_state_dict_config to be private. If device_mesh is used with sharded model/optim state dict, _use_dtensor flag would be set to True and model/optim state dict would return dtensor state_dict. Otherwise, _use_dtensor flag would be set to False and model/optim state dict would return sharded_tensor state_dict.
3) Update _optim_utils.py, _shard_utils.py, and _state_dict_utils.py to add support for HSDP to return 2D DTensor state_dict.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/107533
Approved by: https://github.com/fegin, https://github.com/awgu, https://github.com/wanchaol
There are many communication operations for shardedTensor in the state dict of fsdp. They use the external passed-in pg (or the default pg), which currently supports cuda devices. Before communication, the memory will be moved to cuda, which is implicit (because it is essentially moving data to the memory type required by pg, not the computing device type). Similarly, when users use fsdp on a custom backend, they will pass in a custom pg (which does not support cuda devices), which may cause fsdp to not work properly in some cases. This PR obtains the memory type supported by the pg through _get_pg_default_device during communication, and moves the data to it when needed.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/101533
Approved by: https://github.com/awgu
This adds `FSDPExtensions` to enable TP + FSDP composability. To be agnostic to both `ShardedTensor` and `DistributedTensor`, the design relies on customizable hooks.
Some notes:
- I preferred the `_ext` prefix (short for "extension") over `_param_extension` simply because it is shorter. It should not matter much because it is purely internal facing.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/85039
Approved by: https://github.com/kumpera, https://github.com/fegin
FSDP originally uses `_init_from_local_shards_and_global_metadata()` to create a ShardedTensor for sharded_state_dict(). We have seen some non-trivial overhead if the number of tensors is large. Using `_init_from_local_shards_and_global_metadata ` can significantly reduce the overhead. For a model with ~250 tensors in the state_dict trained with 16 GPUs, the original `sharded_state_dict` takes ~1.7 seconds and this PR reduces the overhead to ~0.6 seconds.
Differential Revision: [D38452170](https://our.internmc.facebook.com/intern/diff/D38452170/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/82911
Approved by: https://github.com/awgu
