Skip to content

Update Benchmarks: Vollo SDK v27.0.1 #15

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
ConorWilliams merged 1 commit into from
Apr 21, 2026
Merged

Update Benchmarks: Vollo SDK v27.0.1 #15

Show file tree
Hide file tree
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
20 changes: 10 additions & 10 deletions benchmarks/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
# Vollo Model Zoo Benchmarks (version 27.0.0)
# Vollo Model Zoo Benchmarks (version 27.0.1)

Compute latency for an approximately 1-million parameter model.

Expand All @@ -8,8 +8,8 @@ Note: These latencies are from a (near cycle-accurate) software model but withou

| Model | Latency (us) | Latency contiguous (us) | Metadata |
| ----- | ------------ | ------------------------ | -------- |
| wavenet | 2.64 | 2.76 | layers=4, blocks=1, hidden=198 |
| tcn | 1.14 | 1.52 | inputs=1, kernel=3, channels=[256, 256, 256] |
| wavenet | 2.60 | 2.70 | layers=4, blocks=1, hidden=198 |
| tcn | 1.12 | 1.51 | inputs=1, kernel=3, channels=[256, 256, 256] |
| ssm | 0.81 | 0.89 | dim=576, hidden=448 |
| slp | 0.86 | 0.87 | input=1024, output=1024, activation=ReLU |
| resmlp | 6.83 | 6.83 | dim=160, patches=9, layers=5, activation=ReLU |
Expand All @@ -28,8 +28,8 @@ Note: These latencies are from a (near cycle-accurate) software model but withou

| Model | Latency (us) | Latency contiguous (us) | Metadata |
| ----- | ------------ | ------------------------ | -------- |
| wavenet | 2.90 | 3.04 | layers=4, blocks=1, hidden=198 |
| tcn | 1.25 | 1.67 | inputs=1, kernel=3, channels=[256, 256, 256] |
| wavenet | 2.86 | 2.97 | layers=4, blocks=1, hidden=198 |
| tcn | 1.24 | 1.66 | inputs=1, kernel=3, channels=[256, 256, 256] |
| ssm | 0.89 | 0.98 | dim=576, hidden=448 |
| slp | 0.95 | 0.95 | input=1024, output=1024, activation=ReLU |
| resmlp | 7.51 | 7.52 | dim=160, patches=9, layers=5, activation=ReLU |
Expand All @@ -48,8 +48,8 @@ Note: These latencies are from a (near cycle-accurate) software model but withou

| Model | Latency (us) | Latency contiguous (us) | Metadata |
| ----- | ------------ | ------------------------ | -------- |
| wavenet | 3.06 | 3.12 | layers=4, blocks=1, hidden=198 |
| tcn | 1.31 | 1.70 | inputs=1, kernel=3, channels=[256, 256, 256] |
| wavenet | 3.02 | 3.13 | layers=4, blocks=1, hidden=198 |
| tcn | 1.30 | 1.70 | inputs=1, kernel=3, channels=[256, 256, 256] |
| ssm | 0.87 | 0.96 | dim=576, hidden=448 |
| slp | 0.91 | 0.91 | input=1024, output=1024, activation=ReLU |
| resmlp | 12.40 | 12.40 | dim=160, patches=9, layers=5, activation=ReLU |
Expand Down Expand Up @@ -82,14 +82,14 @@ Note: These latencies are from a (near cycle-accurate) software model but withou
| lstm | 0.59 | 0.61 | layers=2, hidden_size=250 |
| gru | 0.70 | 1.09 | fp32=False, input=512, hidden=384, layers=1 |
| ffn-swiglu | 1.23 | 1.24 | dim=288, hidden=1152, activation=SwiGLU, fused=True |
| cnn | 1.61 | 3.07 | layers=4, channels=64, kernel_size=64 |
| cnn | 1.61 | 3.06 | layers=4, channels=64, kernel_size=64 |

## Configuration: IA-420f

| Model | Latency (us) | Latency contiguous (us) | Metadata |
| ----- | ------------ | ------------------------ | -------- |
| wavenet | 3.06 | 3.12 | layers=4, blocks=1, hidden=198 |
| tcn | 1.31 | 1.70 | inputs=1, kernel=3, channels=[256, 256, 256] |
| wavenet | 3.02 | 3.13 | layers=4, blocks=1, hidden=198 |
| tcn | 1.30 | 1.70 | inputs=1, kernel=3, channels=[256, 256, 256] |
| ssm | 0.87 | 0.96 | dim=576, hidden=448 |
| slp | 0.91 | 0.91 | input=1024, output=1024, activation=ReLU |
| resmlp | 12.40 | 12.40 | dim=160, patches=9, layers=5, activation=ReLU |
Expand Down
Loading
Loading