Skip to content

Fix input_dist for when it has to handle KVZCH and non-KVZCH tables together for row-wise sharding #3444

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
+9 −0
Open

Fix input_dist for when it has to handle KVZCH and non-KVZCH tables together for row-wise sharding #3444

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 9 additions & 0 deletions torchrec/distributed/sharding/rw_sharding.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -301,6 +301,11 @@ def _get_writable_feature_hash_sizes(self) -> List[int]:
return feature_hash_sizes

def _get_virtual_table_feature_num_buckets(self) -> List[int]:
"""
Returns the number of buckets for each KVZCH feature in the GroupedEmbeddingConfigs.
If a feature is not a KVZCH feature, the list will have world_size for that feature's corresponding position.
This is needed as KVZCH features have to be processed for input_dist with non-KVZCH features.
"""
feature_num_buckets: List[int] = []
for group_config in self._grouped_embedding_configs:
for embedding_table in group_config.embedding_tables:
Expand All @@ -312,6 +317,10 @@ def _get_virtual_table_feature_num_buckets(self) -> List[int]:
[embedding_table.total_num_buckets]
* embedding_table.num_features()
)
else:
feature_num_buckets.extend(
[self._world_size] * embedding_table.num_features()
)
return feature_num_buckets


Expand Down
Loading