Skip to content
KernelIndex
Search⌘K

Transformers

Hugging Face Transformers · python · Apache-2.0
holds 5 records

Use it

Source available upstream · Apache-2.0View source →

No install recipe recorded for this revision.

interfacepython
Compatibility
measured onNVIDIA H100
declared hardwaredeclared only
architectures—
dtypes

Benchmark evidence

10 measurements across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
Fused linear + KTO lossbf16 · [2, 512, 1024]
NVIDIA H100
3.88ms
#1 of 2
2025-03-03
Fused linear + KTO lossbf16 · [2, 512, 1024]
NVIDIA H100
6.20ms
#2 of 2
2025-03-03
Fused linear + KTO lossbf16 · [4, 512, 1024]
NVIDIA H100
7.21ms
#1 of 2
2025-03-03
Fused linear + KTO lossbf16 · [4, 512, 1024]
NVIDIA H100
11.2ms
#2 of 2
2025-03-03
Fused linear + KTO lossbf16 · [8, 512, 1024]
NVIDIA H100
13.8ms
#1 of 2
2025-03-03
Fused linear + KTO lossbf16 · [8, 512, 1024]
NVIDIA H100
21.2ms
#2 of 2
2025-03-03
Fused linear + KTO lossbf16 · [16, 512, 1024]
NVIDIA H100
27.1ms
#1 of 2
2025-03-03
Fused linear + KTO lossbf16 · [16, 512, 1024]
NVIDIA H100
41.6ms
#2 of 2
2025-03-03
Fused linear + KTO lossbf16 · [32, 512, 1024]
NVIDIA H100
54.1ms
#1 of 2
2025-03-03
Fused linear + KTO lossbf16 · [32, 512, 1024]
NVIDIA H100
81.6ms
#2 of 2
2025-03-03

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:df6427ef8ed7d7d4f56c47ca0252c3ea15dc4283d04093de0ff6853aff06c2d0
license declaredApache-2.0
license concludedApache-2.0
imported2026-08-20

Best evidence level for this revision: reported

JSON