Skip to content
KernelIndex
Search⌘K
Sign in

PyTorch

python · BSD-3-Clause

Use it

Source available upstream · BSD-3-ClauseView source →

No install recipe recorded for this revision.

interfacepython
Compatibility
measured onNVIDIA H100
declared hardwaredeclared only
architectures—
dtypes

Benchmark evidence

36 measurements across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 2048]
NVIDIA H100
23.2µs
#2 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 2048]
NVIDIA H100
57.8µs
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 4096]
NVIDIA H100
60.4µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 4096]
NVIDIA H100
105.3µs
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 8192]
NVIDIA H100
113.7µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 2048]
NVIDIA H100
146.9µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 2048]
NVIDIA H100
172.6µs
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 2048]
NVIDIA H100
172.7µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 2048]
NVIDIA H100
178.0µs
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 8192]
NVIDIA H100
197.6µs
#4 of 4
2026-07-26
Run →
Show all 36 measurements ›
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 16384]
NVIDIA H100
220.2µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 4096]
NVIDIA H100
318.0µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 4096]
NVIDIA H100
318.4µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 4096]
NVIDIA H100
338.4µs
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 4096]
NVIDIA H100
338.5µs
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 16384]
NVIDIA H100
383.0µs
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 32768]
NVIDIA H100
432.9µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 8192]
NVIDIA H100
602.7µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 8192]
NVIDIA H100
603.3µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 8192]
NVIDIA H100
656.9µs
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 8192]
NVIDIA H100
657.0µs
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 32768]
NVIDIA H100
750.3µs
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 65536]
NVIDIA H100
858.5µs
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 16384]
NVIDIA H100
1.17ms
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 16384]
NVIDIA H100
1.17ms
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 16384]
NVIDIA H100
1.29ms
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 16384]
NVIDIA H100
1.29ms
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 65536]
NVIDIA H100
1.50ms
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 32768]
NVIDIA H100
2.30ms
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 32768]
NVIDIA H100
2.30ms
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 32768]
NVIDIA H100
2.55ms
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 32768]
NVIDIA H100
2.55ms
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 65536]
NVIDIA H100
4.57ms
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 65536]
NVIDIA H100
4.57ms
#3 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 65536]
NVIDIA H100
5.08ms
#4 of 4
2026-07-26
Run →
SwiGLU activation, Megatron layoutbf16 · [2048, 4, 65536]
NVIDIA H100
5.08ms
#4 of 4
2026-07-26
Run →

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:ebdec3855511a3b6ae7df50fa27a70da8c06b0eb5d8026a2fe27779438c068bb
license declaredBSD-3-Clause
license concludedBSD-3-Clause
imported2026-08-20

Best evidence level for this revision: reported

JSON