Skip to content

Commit 4db55b3

Browse files
committed
docs(bench): record current Ubuntu benchmark evidence
1 parent 9d0a626 commit 4db55b3

2 files changed

Lines changed: 18 additions & 10 deletions

File tree

‎benchmark_results.txt‎

Lines changed: 9 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1,16 +1,19 @@
11
# DINOv2 inference benchmark (forward pass wall time, ms; lower is better).
2-
# Actual artifact from GitHub Actions run 35629935462:
3-
# https://github.com/espetro/dinov2.cpp/actions/runs/35629935462
2+
# Actual artifact from GitHub Actions run 35630688827:
3+
# https://github.com/espetro/dinov2.cpp/actions/runs/35630688827
44
# Branch: feat/batched-inference
5-
# Commit: 50d5dee10da4db7b7820082da7a3ea0eaab5f346
5+
# Commit: 9d0a6265ae53088e423e4c2417ec8589de181e99
66
# Workflow inputs: variant=small, repeats=5, threads=2
7-
# Artifact: bench-results-small-35629935462/benchmark_results.txt
8-
# Generated by scripts/bench.sh on 2026-09-21T17:08:06Z.
7+
# Artifact: bench-results-small-35630688827/benchmark_results.txt
8+
# Provenance artifact: bench-results-small-35630688827/benchmark_provenance.txt
9+
# Model path: models/dinov2-vit-small-patch14/model.gguf
10+
# Model SHA256: b7ca009aa416f6be85ea95363f4ef13d3c139999f96ca3fd47226849afb712e6
11+
# Generated by scripts/bench.sh on 2026-09-21T17:14:58Z.
912
# Platform: linux-x86_64 on ubuntu-latest; CPU backend; f16 GGUF only.
1013
# Methodology: 1 warmup + 5 timed runs, one 224x224 image (batch=1), 2 OpenMP threads.
1114

1215
| Model | mean (ms) | stddev (ms) | min (ms) | max (ms) | peak RSS (MB) |
1316
|:------|----------:|------------:|---------:|---------:|---------------:|
14-
| small | 212.8 | 9.1 | 204 | 225 | 104 |
17+
| small | 255.0 | 9.1 | 249 | 271 | 104 |
1518

1619
# Multi-platform measurements are deferred. No quantization-specific benchmark was run.

‎docs/benchmarks.md‎

Lines changed: 9 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -5,11 +5,16 @@ The table below contains the actual Ubuntu benchmark artifact produced by
55
measurements are deferred.
66

77
The benchmark ran on `.github/workflows/bench.yml` run
8-
[35629935462](https://github.com/espetro/dinov2.cpp/actions/runs/35629935462)
8+
[35630688827](https://github.com/espetro/dinov2.cpp/actions/runs/35630688827)
99
from `feat/batched-inference` at commit
10-
`50d5dee10da4db7b7820082da7a3ea0eaab5f346`. The workflow inputs were
10+
`9d0a6265ae53088e423e4c2417ec8589de181e99`. The workflow inputs were
1111
`variant=small`, `repeats=5`, and `threads=2`; the uploaded artifact was
12-
`bench-results-small-35629935462/benchmark_results.txt`.
12+
`bench-results-small-35630688827/benchmark_results.txt`, accompanied by
13+
`benchmark_provenance.txt` in the same artifact. The staged model was
14+
`models/dinov2-vit-small-patch14/model.gguf` with SHA-256
15+
`b7ca009aa416f6be85ea95363f4ef13d3c139999f96ca3fd47226849afb712e6`. The
16+
artifact recorded `workflow=bench.yml`, `run_id=35630688827`,
17+
`run_attempt=1`, and the same commit SHA.
1318

1419
## Forward pass wall time (CPU backend)
1520

@@ -18,7 +23,7 @@ and PCA visualization. Numbers in milliseconds (lower is better).
1823

1924
| Platform | Arch | ggml | Model | mean (ms) | stddev (ms) | min (ms) | max (ms) | Peak RSS |
2025
|-----------------|--------|------|--------|----------:|------------:|---------:|---------:|---------:|
21-
| ubuntu-latest | x86_64 | cpu | ViT-S | 212.8 | 9.1 | 204 | 225 | 104 MB |
26+
| ubuntu-latest | x86_64 | cpu | ViT-S | 255.0 | 9.1 | 249 | 271 | 104 MB |
2227

2328
This is the only current platform measurement. Other platform rows remain
2429
deferred, and no quantization-specific benchmark was run.

0 commit comments

Comments
 (0)