Skip to content
KernelIndex
Search⌘K

Liger

Liger-Kernel · triton · BSD-2-Clause
holds 13 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

32 measurements across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
Embedding lookupfp32 · [1024, 768]
NVIDIA A100
42.6µs
#2 of 3
2024-09-03
Embedding lookupfp32 · [2048, 768]
NVIDIA A100
46.7µs
#2 of 3
2024-09-03
Embedding lookupfp32 · [4096, 768]
NVIDIA A100
49.4µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [8192, 768]
NVIDIA A100
55.6µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [16384, 768]
NVIDIA A100
61.5µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [32768, 768]
NVIDIA A100
65.2µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [65536, 768]
NVIDIA A100
67.8µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [131072, 768]
NVIDIA A100
70.9µs
#2 of 3
2024-09-03
Embedding lookupfp32 · [1024, 4096]
NVIDIA A100
189.6µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [2048, 4096]
NVIDIA A100
213.0µs
#1 of 3
2024-09-03
Show all 32 measurements ›
Embedding lookupfp32 · [4096, 4096]
NVIDIA A100
236.7µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [8192, 4096]
NVIDIA A100
263.4µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [16384, 4096]
NVIDIA A100
285.1µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [32768, 4096]
NVIDIA A100
304.6µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [65536, 4096]
NVIDIA A100
315.7µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [131072, 4096]
NVIDIA A100
319.9µs
#1 of 3
2024-09-03
Embedding lookupfp32 · [1024, 768]
NVIDIA A100
445.2µs
#2 of 3
2024-09-03
Embedding lookupfp32 · [2048, 768]
NVIDIA A100
462.1µs
#2 of 3
2024-09-03
Embedding lookupfp32 · [4096, 768]
NVIDIA A100
491.4µs
#2 of 3
2024-09-03
Embedding lookupfp32 · [8192, 768]
NVIDIA A100
552.8µs
#2 of 3
2024-09-03
Embedding lookupfp32 · [16384, 768]
NVIDIA A100
635.0µs
#2 of 3
2024-09-03
Embedding lookupfp32 · [32768, 768]
NVIDIA A100
771.1µs
#2 of 3
2024-09-03
Embedding lookupfp32 · [65536, 768]
NVIDIA A100
1.00ms
#2 of 3
2024-09-03
Embedding lookupfp32 · [131072, 768]
NVIDIA A100
1.45ms
#2 of 3
2024-09-03
Embedding lookupfp32 · [1024, 4096]
NVIDIA A100
2.18ms
#3 of 3
2024-09-03
Embedding lookupfp32 · [2048, 4096]
NVIDIA A100
2.29ms
#3 of 3
2024-09-03
Embedding lookupfp32 · [4096, 4096]
NVIDIA A100
2.48ms
#3 of 3
2024-09-03
Embedding lookupfp32 · [8192, 4096]
NVIDIA A100
2.74ms
#3 of 3
2024-09-03
Embedding lookupfp32 · [16384, 4096]
NVIDIA A100
3.12ms
#3 of 3
2024-09-03
Embedding lookupfp32 · [32768, 4096]
NVIDIA A100
3.78ms
#2 of 3
2024-09-03
Embedding lookupfp32 · [65536, 4096]
NVIDIA A100
4.99ms
#2 of 3
2024-09-03
Embedding lookupfp32 · [131072, 4096]
NVIDIA A100
7.38ms
#2 of 3
2024-09-03

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:6f0b2f81fb53464a0c7be7d82ecc3b4293d50ea1e9fe7e5ee6d64d67c8a43e85
license declaredBSD-2-Clause
license concludedBSD-2-Clause
imported2026-08-18

Best evidence level for this revision: reported

JSON