Skip to content
KernelIndex
Search⌘K
Sign in

Liger fused

Liger-Kernel · triton · BSD-2-Clause
holds 16 records

Use it

Installable · pip · pinned to the measured build · BSD-2-ClauseView source →

pip install "liger-kernel==0.7.0"

Pinned: this command installs the newest release the evidence below measured.

interfacetriton
Compatibility
measured onNVIDIA H100
declared hardwaredeclared only
architectures—
dtypes

Benchmark evidence

36 measurements across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
Fused add + RMSNormfp32 · [2048, 1024]
NVIDIA H100
20.5µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 2048]
NVIDIA H100
28.2µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 1024]
NVIDIA H100
46.1µs
#3 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 2048]
NVIDIA H100
47.2µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 4096]
NVIDIA H100
49.5µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 8192]
NVIDIA H100
94.6µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 4096]
NVIDIA H100
95.8µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 4096]
NVIDIA H100
125.2µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 2048]
NVIDIA H100
125.3µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 1024]
NVIDIA H100
125.9µs
#2 of 3
2026-04-07
Run →
Show all 36 measurements ›
Fused add + RMSNormfp32 · [2048, 1024]
NVIDIA H100
139.2µs
#3 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 2048]
NVIDIA H100
140.2µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 8192]
NVIDIA H100
154.5µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 8192]
NVIDIA H100
177.1µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 16384]
NVIDIA H100
183.8µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 4096]
NVIDIA H100
211.7µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 1024]
NVIDIA H100
238.8µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 2048]
NVIDIA H100
277.0µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 4096]
NVIDIA H100
277.0µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 8192]
NVIDIA H100
290.6µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 16384]
NVIDIA H100
300.5µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 16384]
NVIDIA H100
340.2µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 2048]
NVIDIA H100
355.9µs
#3 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 1024]
NVIDIA H100
356.1µs
#3 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 4096]
NVIDIA H100
361.6µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 8192]
NVIDIA H100
385.7µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 32768]
NVIDIA H100
433.8µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 16384]
NVIDIA H100
482.3µs
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 8192]
NVIDIA H100
559.3µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 32768]
NVIDIA H100
683.3µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 16384]
NVIDIA H100
754.0µs
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 16384]
NVIDIA H100
1.09ms
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 32768]
NVIDIA H100
1.12ms
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 32768]
NVIDIA H100
1.55ms
#1 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 32768]
NVIDIA H100
2.01ms
#2 of 3
2026-04-07
Run →
Fused add + RMSNormfp32 · [2048, 32768]
NVIDIA H100
2.69ms
#2 of 3
2026-04-07
Run →

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:9b417b271806017a7e78c087a31c17a4633f00e5df9c8b3c227d337c5aea8505
license declaredBSD-2-Clause
license concludedBSD-2-Clause
imported2026-08-18

Best evidence level for this revision: reported

JSON