Skip to content
KernelIndex
Search⌘K
Sign in

Liger

Liger-Kernel · triton · BSD-2-Clause
holds 20 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

24 measurements across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
ReLU squaredbf16 · [4096, 128]
NVIDIA H100
7.33µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 256]
NVIDIA H100
7.39µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 128]
NVIDIA H100
7.94µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 256]
NVIDIA H100
8.35µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 512]
NVIDIA H100
8.48µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 512]
NVIDIA H100
10.6µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 1024]
NVIDIA H100
11.3µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 1024]
NVIDIA H100
14.7µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 2048]
NVIDIA H100
17.1µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 2048]
NVIDIA H100
23.3µs
#1 of 2
2026-03-27
Run →
Show all 24 measurements ›
ReLU squaredbf16 · [4096, 4096]
NVIDIA H100
27.9µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 4096]
NVIDIA H100
39.4µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 8192]
NVIDIA H100
50.2µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 8192]
NVIDIA H100
72.4µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 2048]
NVIDIA H100
75.6µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 1024]
NVIDIA H100
77.4µs
#2 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 512]
NVIDIA H100
78.9µs
#2 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 4096]
NVIDIA H100
87.1µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 256]
NVIDIA H100
88.4µs
#2 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 16384]
NVIDIA H100
94.4µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 8192]
NVIDIA H100
119.4µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 128]
NVIDIA H100
124.1µs
#2 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 16384]
NVIDIA H100
137.1µs
#1 of 2
2026-03-27
Run →
ReLU squaredbf16 · [4096, 16384]
NVIDIA H100
229.0µs
#1 of 2
2026-03-27
Run →

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:22a7fa375bb88a22a8cfc467d61a872f5044d782d848daba169df1a6f9186b8d
license declaredBSD-2-Clause
license concludedBSD-2-Clause
imported2026-08-20

Best evidence level for this revision: reported

JSON