Skip to content
KernelIndex
Search⌘K
Sign in

PyTorch

python · BSD-3-Clause
holds 6 records

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

30 measurements across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
RMSNorm, Megatron layoutbf16 · [4096, 1, 1024]
NVIDIA H100
12.7µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 1024]
NVIDIA H100
12.8µs
#3 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 2048]
NVIDIA H100
19.0µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 2048]
NVIDIA H100
19.0µs
#3 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 4096]
NVIDIA H100
32.5µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 4096]
NVIDIA H100
32.5µs
#2= of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 1024]
NVIDIA H100
74.4µs
#1 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 8192]
NVIDIA H100
74.5µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 8192]
NVIDIA H100
74.5µs
#2= of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 1024]
NVIDIA H100
74.6µs
#1 of 3
2026-06-10
Run →
Show all 30 measurements ›
RMSNorm, Megatron layoutbf16 · [4096, 1, 1024]
NVIDIA H100
74.6µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 1024]
NVIDIA H100
76.1µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 2048]
NVIDIA H100
93.2µs
#1 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 2048]
NVIDIA H100
94.7µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 2048]
NVIDIA H100
108.5µs
#1 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 2048]
NVIDIA H100
108.6µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 16384]
NVIDIA H100
149.2µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 16384]
NVIDIA H100
149.3µs
#3 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 4096]
NVIDIA H100
162.4µs
#1 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 4096]
NVIDIA H100
162.9µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 4096]
NVIDIA H100
195.1µs
#1 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 4096]
NVIDIA H100
195.3µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 8192]
NVIDIA H100
334.1µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 8192]
NVIDIA H100
336.6µs
#3 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 8192]
NVIDIA H100
396.5µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 8192]
NVIDIA H100
396.7µs
#3 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 16384]
NVIDIA H100
642.8µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 16384]
NVIDIA H100
643.5µs
#3 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 16384]
NVIDIA H100
771.7µs
#2 of 3
2026-06-10
Run →
RMSNorm, Megatron layoutbf16 · [4096, 1, 16384]
NVIDIA H100
771.7µs
#3 of 3
2026-06-10
Run →

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:07486a893d05b3338eb1876548053826c12ab08c8afa12bcc829975f75e540c2
license declaredBSD-3-Clause
license concludedBSD-3-Clause
imported2026-08-20

Best evidence level for this revision: reported

JSON