Skip to content

Add Sparsify overhead benchmark #3021

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

Merged
merged 3 commits into from
Sep 25, 2025
Merged

Add Sparsify overhead benchmark #3021

Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
50ac2cc
Summary:
namgyu-youn Sep 17, 2025
f9f2f8d
remove lambda, scale for fair comparison
namgyu-youn Sep 19, 2025
cfbeabf
rename attributes to prevent duplicate naming
namgyu-youn Sep 21, 2025
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
15 changes: 15 additions & 0 deletions benchmarks/benchmark_e2e_fp8_sparse_linear.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,18 @@ def benchmark(num_tokens, hidden_size=8192, intermediate_size=8192):
input_tensor = torch.randn(num_tokens, hidden_size).to(torch.bfloat16).cuda()
fp16_time = benchmark_microseconds(ffn_ref, input_tensor)

# Sparsify-only benchmarks
ao_fast_sparsification_time = benchmark_microseconds(
torch.ops.torchao.sparse24_sm90_sparsify(
input_tensor,
"cutlass",
"srelu",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this should be "identity" here instead

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

please update this :)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the reminder; I missed it.

"largest",
dtype=torch.float8_e4m3fn,
)
)
cusparse_time = benchmark_microseconds(torch._cslt_compress, input_tensor)

# bf16
ffn_clone = (
nn.Sequential(
Expand Down Expand Up @@ -117,7 +129,10 @@ def benchmark(num_tokens, hidden_size=8192, intermediate_size=8192):
"fp8_c_time (us)": fp8_c_time,
"fp8_c_sparse_time (us)": fp8_c_sparse_time,
"fp8_c_activation_sparse_time (us)": fp8_c_activation_sparse_time,
"ao_fast_sparsification_time (us)": ao_fast_sparsification_time,
"cusparse*_compress_time (us)": cusparse_time,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

cusparse_time isnt a good name for this because there is a seperate cusparse library, aside from cusparselt. please use cusparselt here instead

Also looks like theres a typo in the string?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, I didn't know it due to my lack of background.

"speedup": fp8_c_time / fp8_c_activation_sparse_time,
"sparsify_speedup": cusparse_time / ao_fast_sparsification_time,
}


Expand Down
Loading