+

SwiGLU Activation Benchmarks - Aggregated Results (macOS)

+

This document combines benchmark results from SwiGLU activation implementations on macOS.

+

Combined Summary and Visualization

+
+ + + + + + + 2025-12-19T22:44:10.356853 + image/svg+xml + + + Matplotlib v3.10.8, https://matplotlib.org/ + + + + + + + + + + + + + + +
+ +
+
+ +▶ code +▼ output + ▶ uv-logs + | +Cell: combine | 20.76s + | + +Raw +
+ +
+
======================================================================
+LOADING BENCHMARK DATA
+======================================================================
+✓ PyTorch SwiGLU (macOS)        : /Users/runner/work/kernels-benchmarks/kernels-benchmarks/benches/activation/impls/.uvnote/cache/f249c063d33d6e52c05f603fc062dc2bd9327140dcd61028ab7d7192fd88a5c6
+
+  ✓ Found PyTorch SwiGLU (macOS)
+     Path: /Users/runner/work/kernels-benchmarks/kernels-benchmarks/benches/activation/impls/.uvnote/cache/f249c063d33d6e52c05f603fc062dc2bd9327140dcd61028ab7d7192fd88a5c6/activation.jsonl
+
+======================================================================
+Summary: 1 found, 0 skipped, 0 missing
+======================================================================
+
+COMBINED BENCHMARK SUMMARY
+
+impl                     wl                  p50(ms)  ok
+torch_eager_darwin       cpu_T128_D1024         0.18  False
+torch_eager_darwin       cpu_T128_D2048         0.38  False
+torch_eager_darwin       cpu_T128_D768          0.14  False
+torch_eager_darwin       cpu_T256_D1024         0.40  False
+torch_eager_darwin       cpu_T256_D2048         0.98  False
+torch_eager_darwin       cpu_T256_D768          0.33  False
+torch_eager_darwin       cpu_T512_D1024         1.63  False
+torch_eager_darwin       cpu_T512_D2048         3.37  False
+torch_eager_darwin       cpu_T512_D768          0.87  False
+
+GENERATING COMBINED VISUALIZATION
+
+Loaded 9 records
+No valid records found
+✓ Visualization saved as latency.svg
+✓ SVG visualization ready!
+
+ANALYSIS COMPLETE
+Total implementations analyzed: 1
+
+Implementations included:
+  ✓ PyTorch SwiGLU (macOS)
+
+
+
▶ UV Install Logs
+ +
+
Matplotlib is building the font cache; this may take a moment.
+
+

Artifacts:

+latency.svg +
+ + + + + + + 2025-12-19T22:44:10.356853 + image/svg+xml + + + Matplotlib v3.10.8, https://matplotlib.org/ + + + + + + + + + + + + + + +
+
+
+
+