Skip to content
KernelIndex
Search⌘K
Sign in

Transformers

Hugging Face Transformers · python · Apache-2.0

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

42 measurements across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
Fused MoE expert MLPbf16 · [8192, 2048] · experts=8
NVIDIA H100
4.48ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=16
NVIDIA H100
5.75ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=8
NVIDIA H100
5.92ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=32
NVIDIA H100
9.27ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=16
NVIDIA H100
9.96ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=8
NVIDIA H100
10.6ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=16
NVIDIA H100
15.9ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=64
NVIDIA H100
16.7ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=32
NVIDIA H100
21.5ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [512, 2048]
NVIDIA H100
30.2ms
#2 of 2
2026-04-02
Run →
Show all 42 measurements ›
Fused MoE expert MLPbf16 · [256, 2048]
NVIDIA H100
30.3ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [128, 2048]
NVIDIA H100
30.3ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=32
NVIDIA H100
31.2ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [1024, 2048]
NVIDIA H100
31.2ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [4096, 2048]
NVIDIA H100
31.4ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [2048, 2048]
NVIDIA H100
31.5ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=128
NVIDIA H100
31.5ms
#3 of 4
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=128
NVIDIA H100
32.0ms
#4 of 4
2026-04-02
Run →
Fused MoE expert MLPbf16 · [16384, 2048]
NVIDIA H100
33.3ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [32768, 2048]
NVIDIA H100
37.6ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=64
NVIDIA H100
64.0ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=64
NVIDIA H100
81.3ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [128, 2048]
NVIDIA H100
214.2ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [512, 2048]
NVIDIA H100
214.6ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [256, 2048]
NVIDIA H100
214.7ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [1024, 2048]
NVIDIA H100
215.4ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [2048, 2048]
NVIDIA H100
216.5ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [4096, 2048]
NVIDIA H100
218.5ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=128
NVIDIA H100
222.8ms
#3 of 4
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=128
NVIDIA H100
223.5ms
#4 of 4
2026-04-02
Run →
Fused MoE expert MLPbf16 · [16384, 2048]
NVIDIA H100
232.1ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [128, 2048]
NVIDIA H100
244.3ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [512, 2048]
NVIDIA H100
244.7ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [256, 2048]
NVIDIA H100
244.8ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [1024, 2048]
NVIDIA H100
246.5ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [2048, 2048]
NVIDIA H100
248.0ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [32768, 2048]
NVIDIA H100
249.8ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [4096, 2048]
NVIDIA H100
250.2ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=128
NVIDIA H100
255.1ms
#3 of 4
2026-04-02
Run →
Fused MoE expert MLPbf16 · [8192, 2048] · experts=128
NVIDIA H100
255.8ms
#4 of 4
2026-04-02
Run →
Fused MoE expert MLPbf16 · [16384, 2048]
NVIDIA H100
266.1ms
#2 of 2
2026-04-02
Run →
Fused MoE expert MLPbf16 · [32768, 2048]
NVIDIA H100
287.9ms
#2 of 2
2026-04-02
Run →

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:205272a6edab93cf893ad6d2fa2cc9786bfbfa97a1a7fdab0ed7f41768333593
license declaredApache-2.0
license concludedApache-2.0
imported2026-08-20

Best evidence level for this revision: reported

JSON