Liger
Liger-Kernel · triton · BSD-2-Clause
holds 18 records
Use it
Installable · pip · pinned to the measured build · BSD-2-ClauseView source →
pip install "liger-kernel==0.8.0"Pinned: this command installs the newest release the evidence below measured.
interfacetriton
Compatibility
measured onNVIDIA H100
declared hardwaredeclared only
architectures—
dtypes
Benchmark evidence
18 measurements across 1 GPU, fastest first.
Operation / workload
Hardware
Latency
Rank
Observed
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 4096]
NVIDIA H100
138.6µs
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 8192]
NVIDIA H100
262.6µs
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 4096]
NVIDIA H100
349.4µs
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 4096]
NVIDIA H100
349.8µs
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 16384]
NVIDIA H100
415.8µs
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 8192]
NVIDIA H100
585.4µs
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 8192]
NVIDIA H100
585.5µs
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 32768]
NVIDIA H100
809.8µs
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 16384]
NVIDIA H100
1.05ms
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 16384]
NVIDIA H100
1.05ms
#1 of 3
2026-06-15
Show all 18 measurements ›Showing all 18 measurements ⌄
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 65536]
NVIDIA H100
1.53ms
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 32768]
NVIDIA H100
2.06ms
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 32768]
NVIDIA H100
2.06ms
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 131072]
NVIDIA H100
2.97ms
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 65536]
NVIDIA H100
4.02ms
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 65536]
NVIDIA H100
4.02ms
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 131072]
NVIDIA H100
7.96ms
#1 of 3
2026-06-15
Vocab-parallel cross-entropy, Megatron layoutbf16 · [2048, 4, 131072]
NVIDIA H100
7.96ms
#1 of 3
2026-06-15
Reported · How evidence levels are derived →
Source and license
sourceavailable
revision digestsha256:f8317f62e3f6564e5d132fe59b9d9d97b424c5b6e9f91d92915e8d8007681337
license declaredBSD-2-Clause
license concludedBSD-2-Clause
imported2026-08-20
Best evidence level for this revision: reported
JSON