add stride into KJT pytree #2587
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
facebook-github-bot
added
the
CLA Signed
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66400821 |
TroyGarden
force-pushed
the
export-D66400821
branch
from
8b5124f to
534df45
Compare
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Dec 12, 2024
Summary:
# context
* Previously for a KJT, only the following fields and `_keys` are stored in the pytree flatten specs. All other arguments/parameters would be derived accordingly.
```
_fields = [
"_values",
"_weights",
"_lengths",
"_offsets",
]
```
* Particularly, the `stride` (int) of a KJT, which represents the `batch_size`, is computed by `_maybe_compute_stride_kjt`:
```
def _maybe_compute_stride_kjt(
keys: List[str],
stride: Optional[int],
lengths: Optional[torch.Tensor],
offsets: Optional[torch.Tensor],
stride_per_key_per_rank: Optional[List[List[int]]],
) -> int:
if stride is None:
if len(keys) == 0:
stride = 0
elif stride_per_key_per_rank is not None and len(stride_per_key_per_rank) > 0:
stride = max([sum(s) for s in stride_per_key_per_rank])
elif offsets is not None and offsets.numel() > 0:
stride = (offsets.numel() - 1) // len(keys)
elif lengths is not None:
stride = lengths.numel() // len(keys)
else:
stride = 0
return stride
```
* The previously stored pytree flatten specs are enough if the `batch_size` is static, however, this no longer holds true in a variable batch size scenario, where the `stride_per_key_per_rank` is not `None`.
* An example is that with `dedup_ebc`, where the actual batch_size is variable (depending on the dedup data), but the output of the ebc should always be the **true** `stride` (static).
* During ir_export, the output shape will be calculated from `kjt.stride()` function, which would be incorrect if the pytree specs only contains the `keys`.
* This diff adds the `stride` into the KJT pytree flatten/unflatten functions so that a fakified KJT would have the correct stride value.
Differential Revision: D66400821
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66400821 |
1 similar comment
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66400821 |
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Jun 18, 2025
Summary: Pull Request resolved: meta-pytorch#2587 # context * Previously for a KJT, only the following fields and `_keys` are stored in the pytree flatten specs. All other arguments/parameters would be derived accordingly. ``` _fields = [ "_values", "_weights", "_lengths", "_offsets", ] ``` * Particularly, the `stride` (int) of a KJT, which represents the `batch_size`, is computed by `_maybe_compute_stride_kjt`: ``` def _maybe_compute_stride_kjt( keys: List[str], stride: Optional[int], lengths: Optional[torch.Tensor], offsets: Optional[torch.Tensor], stride_per_key_per_rank: Optional[List[List[int]]], ) -> int: if stride is None: if len(keys) == 0: stride = 0 elif stride_per_key_per_rank is not None and len(stride_per_key_per_rank) > 0: stride = max([sum(s) for s in stride_per_key_per_rank]) elif offsets is not None and offsets.numel() > 0: stride = (offsets.numel() - 1) // len(keys) elif lengths is not None: stride = lengths.numel() // len(keys) else: stride = 0 return stride ``` * The previously stored pytree flatten specs are enough if the `batch_size` is static, however, this no longer holds true in a variable batch size scenario, where the `stride_per_key_per_rank` is not `None`. * An example is that with `dedup_ebc`, where the actual batch_size is variable (depending on the dedup data), but the output of the ebc should always be the **true** `stride` (static). * During ir_export, the output shape will be calculated from `kjt.stride()` function, which would be incorrect if the pytree specs only contains the `keys`. * This diff adds the `stride` into the KJT pytree flatten/unflatten functions so that a fakified KJT would have the correct stride value. Reviewed By: PaulZhang12 Differential Revision: D66400821
TroyGarden
force-pushed
the
export-D66400821
branch
from
534df45 to
92d1bf8
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66400821 |
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Jun 18, 2025
Summary: Pull Request resolved: meta-pytorch#2587 # context * Previously for a KJT, only the following fields and `_keys` are stored in the pytree flatten specs. All other arguments/parameters would be derived accordingly. ``` _fields = [ "_values", "_weights", "_lengths", "_offsets", ] ``` * Particularly, the `stride` (int) of a KJT, which represents the `batch_size`, is computed by `_maybe_compute_stride_kjt`: ``` def _maybe_compute_stride_kjt( keys: List[str], stride: Optional[int], lengths: Optional[torch.Tensor], offsets: Optional[torch.Tensor], stride_per_key_per_rank: Optional[List[List[int]]], ) -> int: if stride is None: if len(keys) == 0: stride = 0 elif stride_per_key_per_rank is not None and len(stride_per_key_per_rank) > 0: stride = max([sum(s) for s in stride_per_key_per_rank]) elif offsets is not None and offsets.numel() > 0: stride = (offsets.numel() - 1) // len(keys) elif lengths is not None: stride = lengths.numel() // len(keys) else: stride = 0 return stride ``` * The previously stored pytree flatten specs are enough if the `batch_size` is static, however, this no longer holds true in a variable batch size scenario, where the `stride_per_key_per_rank` is not `None`. * An example is that with `dedup_ebc`, where the actual batch_size is variable (depending on the dedup data), but the output of the ebc should always be the **true** `stride` (static). * During ir_export, the output shape will be calculated from `kjt.stride()` function, which would be incorrect if the pytree specs only contains the `keys`. * This diff adds the `stride` into the KJT pytree flatten/unflatten functions so that a fakified KJT would have the correct stride value. Reviewed By: PaulZhang12 Differential Revision: D66400821
TroyGarden
force-pushed
the
export-D66400821
branch
from
92d1bf8 to
ca0df12
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66400821 |
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Jun 18, 2025
Summary: Pull Request resolved: meta-pytorch#2587 # context * Previously for a KJT, only the following fields and `_keys` are stored in the pytree flatten specs. All other arguments/parameters would be derived accordingly. ``` _fields = [ "_values", "_weights", "_lengths", "_offsets", ] ``` * Particularly, the `stride` (int) of a KJT, which represents the `batch_size`, is computed by `_maybe_compute_stride_kjt`: ``` def _maybe_compute_stride_kjt( keys: List[str], stride: Optional[int], lengths: Optional[torch.Tensor], offsets: Optional[torch.Tensor], stride_per_key_per_rank: Optional[List[List[int]]], ) -> int: if stride is None: if len(keys) == 0: stride = 0 elif stride_per_key_per_rank is not None and len(stride_per_key_per_rank) > 0: stride = max([sum(s) for s in stride_per_key_per_rank]) elif offsets is not None and offsets.numel() > 0: stride = (offsets.numel() - 1) // len(keys) elif lengths is not None: stride = lengths.numel() // len(keys) else: stride = 0 return stride ``` * The previously stored pytree flatten specs are enough if the `batch_size` is static, however, this no longer holds true in a variable batch size scenario, where the `stride_per_key_per_rank` is not `None`. * An example is that with `dedup_ebc`, where the actual batch_size is variable (depending on the dedup data), but the output of the ebc should always be the **true** `stride` (static). * During ir_export, the output shape will be calculated from `kjt.stride()` function, which would be incorrect if the pytree specs only contains the `keys`. * This diff adds the `stride` into the KJT pytree flatten/unflatten functions so that a fakified KJT would have the correct stride value. Reviewed By: PaulZhang12 Differential Revision: D66400821
TroyGarden
force-pushed
the
export-D66400821
branch
from
ca0df12 to
b714b75
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66400821 |
TroyGarden
force-pushed
the
export-D66400821
branch
from
b714b75 to
3c8f53a
Compare
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Jun 18, 2025
Summary: Pull Request resolved: meta-pytorch#2587 # context * Previously for a KJT, only the following fields and `_keys` are stored in the pytree flatten specs. All other arguments/parameters would be derived accordingly. ``` _fields = [ "_values", "_weights", "_lengths", "_offsets", ] ``` * Particularly, the `stride` (int) of a KJT, which represents the `batch_size`, is computed by `_maybe_compute_stride_kjt`: ``` def _maybe_compute_stride_kjt( keys: List[str], stride: Optional[int], lengths: Optional[torch.Tensor], offsets: Optional[torch.Tensor], stride_per_key_per_rank: Optional[List[List[int]]], ) -> int: if stride is None: if len(keys) == 0: stride = 0 elif stride_per_key_per_rank is not None and len(stride_per_key_per_rank) > 0: stride = max([sum(s) for s in stride_per_key_per_rank]) elif offsets is not None and offsets.numel() > 0: stride = (offsets.numel() - 1) // len(keys) elif lengths is not None: stride = lengths.numel() // len(keys) else: stride = 0 return stride ``` * The previously stored pytree flatten specs are enough if the `batch_size` is static, however, this no longer holds true in a variable batch size scenario, where the `stride_per_key_per_rank` is not `None`. * An example is that with `dedup_ebc`, where the actual batch_size is variable (depending on the dedup data), but the output of the ebc should always be the **true** `stride` (static). * During ir_export, the output shape will be calculated from `kjt.stride()` function, which would be incorrect if the pytree specs only contains the `keys`. * This diff adds the `stride` into the KJT pytree flatten/unflatten functions so that a fakified KJT would have the correct stride value. Reviewed By: PaulZhang12 Differential Revision: D66400821
Summary: Pull Request resolved: meta-pytorch#2587 # context * Previously for a KJT, only the following fields and `_keys` are stored in the pytree flatten specs. All other arguments/parameters would be derived accordingly. ``` _fields = [ "_values", "_weights", "_lengths", "_offsets", ] ``` * Particularly, the `stride` (int) of a KJT, which represents the `batch_size`, is computed by `_maybe_compute_stride_kjt`: ``` def _maybe_compute_stride_kjt( keys: List[str], stride: Optional[int], lengths: Optional[torch.Tensor], offsets: Optional[torch.Tensor], stride_per_key_per_rank: Optional[List[List[int]]], ) -> int: if stride is None: if len(keys) == 0: stride = 0 elif stride_per_key_per_rank is not None and len(stride_per_key_per_rank) > 0: stride = max([sum(s) for s in stride_per_key_per_rank]) elif offsets is not None and offsets.numel() > 0: stride = (offsets.numel() - 1) // len(keys) elif lengths is not None: stride = lengths.numel() // len(keys) else: stride = 0 return stride ``` * The previously stored pytree flatten specs are enough if the `batch_size` is static, however, this no longer holds true in a variable batch size scenario, where the `stride_per_key_per_rank` is not `None`. * An example is that with `dedup_ebc`, where the actual batch_size is variable (depending on the dedup data), but the output of the ebc should always be the **true** `stride` (static). * During ir_export, the output shape will be calculated from `kjt.stride()` function, which would be incorrect if the pytree specs only contains the `keys`. * This diff adds the `stride` into the KJT pytree flatten/unflatten functions so that a fakified KJT would have the correct stride value. Reviewed By: PaulZhang12 Differential Revision: D66400821
TroyGarden
force-pushed
the
export-D66400821
branch
from
3c8f53a to
b7bb028
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D66400821 |
TroyGarden
added a commit
to TroyGarden/torchrec
that referenced
this pull request
Jun 18, 2025
Summary: Pull Request resolved: meta-pytorch#2587 # context * Previously for a KJT, only the following fields and `_keys` are stored in the pytree flatten specs. All other arguments/parameters would be derived accordingly. ``` _fields = [ "_values", "_weights", "_lengths", "_offsets", ] ``` * Particularly, the `stride` (int) of a KJT, which represents the `batch_size`, is computed by `_maybe_compute_stride_kjt`: ``` def _maybe_compute_stride_kjt( keys: List[str], stride: Optional[int], lengths: Optional[torch.Tensor], offsets: Optional[torch.Tensor], stride_per_key_per_rank: Optional[List[List[int]]], ) -> int: if stride is None: if len(keys) == 0: stride = 0 elif stride_per_key_per_rank is not None and len(stride_per_key_per_rank) > 0: stride = max([sum(s) for s in stride_per_key_per_rank]) elif offsets is not None and offsets.numel() > 0: stride = (offsets.numel() - 1) // len(keys) elif lengths is not None: stride = lengths.numel() // len(keys) else: stride = 0 return stride ``` * The previously stored pytree flatten specs are enough if the `batch_size` is static, however, this no longer holds true in a variable batch size scenario, where the `stride_per_key_per_rank` is not `None`. * An example is that with `dedup_ebc`, where the actual batch_size is variable (depending on the dedup data), but the output of the ebc should always be the **true** `stride` (static). * During ir_export, the output shape will be calculated from `kjt.stride()` function, which would be incorrect if the pytree specs only contains the `keys`. * This diff adds the `stride` into the KJT pytree flatten/unflatten functions so that a fakified KJT would have the correct stride value. Differential Revision: D66400821 Reviewed By: PaulZhang12
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
context
_keysare stored in the pytree flatten specs. All other arguments/parameters would be derived accordingly.stride(int) of a KJT, which represents thebatch_size, is computed by_maybe_compute_stride_kjt:batch_sizeis static, however, this no longer holds true in a variable batch size scenario, where thestride_per_key_per_rankis notNone.dedup_ebc, where the actual batch_size is variable (depending on the dedup data), but the output of the ebc should always be the truestride(static).kjt.stride()function, which would be incorrect if the pytree specs only contains thekeys.strideinto the KJT pytree flatten/unflatten functions so that a fakified KJT would have the correct stride value.Differential Revision: D66400821