Skip to content

Fix the int4 table batched embedding benchmark with mixed dim #609

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

Closed
wants to merge 1 commit into from
Closed

Fix the int4 table batched embedding benchmark with mixed dim #609

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
6 changes: 4 additions & 2 deletions fbgemm_gpu/bench/split_table_batched_embeddings_benchmark.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -791,7 +791,8 @@ def cpu( # noqa C901
feature_requires_grad = None
if mixed:
Ds = [
div_round_up(np.random.randint(low=int(0.5 * D), high=int(1.5 * D)), 4)
# int4 table batched emb op can only handle mixed D where D is multiple of 8
div_round_up(np.random.randint(low=int(0.5 * D), high=int(1.5 * D)), 8)
for _ in range(T)
]
D = np.average(Ds)
Expand Down Expand Up @@ -905,8 +906,9 @@ def int4_device( # noqa C901
else:
feature_requires_grad = None
if mixed:
# int4 table batched emb op can only handle mixed D where D is multiple of 8
Ds = [
div_round_up(np.random.randint(low=int(0.5 * D), high=int(1.5 * D)), 4)
div_round_up(np.random.randint(low=int(0.5 * D), high=int(1.5 * D)), 8)
for _ in range(T)
]
D = np.average(Ds)
Expand Down