Skip to content
KernelIndex
Search⌘K
Sign in

Liger

Liger-Kernel · triton · BSD-2-Clause
holds 21 records

Use it

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

pip install "liger-kernel==0.2.1"

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

interfacetriton
Compatibility
measured onNVIDIA A100
declared hardwaredeclared only
architectures—
dtypes

Benchmark evidence

24 measurements across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
Rotary position embeddingbf16 · [1, 8, 2048, 16]
NVIDIA A100
11.4µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 64]
NVIDIA A100
20.9µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 1024, 256]
NVIDIA A100
34.4µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 256]
NVIDIA A100
58.9µs
#1 of 4
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 256]
NVIDIA A100
59.5µs
#2 of 4
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 4096, 256]
NVIDIA A100
109.0µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 256]
NVIDIA A100
116.1µs
#1 of 4
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 16]
NVIDIA A100
122.3µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 64]
NVIDIA A100
123.4µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 4096, 256]
NVIDIA A100
124.1µs
#1 of 2
2024-09-03
Run →
Show all 24 measurements ›
Rotary position embeddingbf16 · [1, 8, 1024, 256]
NVIDIA A100
126.0µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 256]
NVIDIA A100
128.1µs
#2 of 4
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 8192, 256]
NVIDIA A100
209.3µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 8192, 256]
NVIDIA A100
210.0µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 4096, 256]
NVIDIA A100
265.1µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 256]
NVIDIA A100
265.5µs
#1 of 4
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 1024, 256]
NVIDIA A100
268.3µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 16]
NVIDIA A100
274.0µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 256]
NVIDIA A100
276.0µs
#2 of 4
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 2048, 64]
NVIDIA A100
281.5µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 16384, 256]
NVIDIA A100
410.2µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 16384, 256]
NVIDIA A100
410.5µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 8192, 256]
NVIDIA A100
415.9µs
#1 of 2
2024-09-03
Run →
Rotary position embeddingbf16 · [1, 8, 16384, 256]
NVIDIA A100
816.7µs
#1 of 2
2024-09-03
Run →

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:51eb16de033f2219f67b768ae0f8b919b4f1af7bd11ee0c19e51fe8df440fa54
license declaredBSD-2-Clause
license concludedBSD-2-Clause
imported2026-08-18

Best evidence level for this revision: reported

JSON