Skip to content
KernelIndex
Search⌘K

PyTorch

python · BSD-3-Clause
holds 14 records

Use it

Source available upstream · BSD-3-ClauseView source →

No install recipe recorded for this revision.

interfacepython
Compatibility
measured onNVIDIA GeForce RTX 3090
declared hardwaredeclared only
architectures—
dtypes

Benchmark evidence

24 measurements across 1 GPU, fastest first.

Operation / workload
Hardware
Latency
Rank
Observed
Softmaxfp32 · [2048, 128]
NVIDIA GeForce RTX 3090
6.14µs
#1 of 2
2025-04-30
Softmaxbf16 · [2048, 128]
NVIDIA GeForce RTX 3090
6.14µs
#1 of 2
2025-04-30
Softmaxbf16 · [2048, 256]
NVIDIA GeForce RTX 3090
6.21µs
#1 of 2
2025-04-30
Softmaxfp32 · [2048, 256]
NVIDIA GeForce RTX 3090
8.19µs
#1 of 2
2025-04-30
Softmaxbf16 · [2048, 512]
NVIDIA GeForce RTX 3090
8.38µs
#2 of 2
2025-04-30
Softmaxbf16 · [2048, 128]
NVIDIA GeForce RTX 3090
11.3µs
#1 of 2
2025-04-30
Softmaxfp32 · [2048, 512]
NVIDIA GeForce RTX 3090
12.3µs
#1 of 2
2025-04-30
Softmaxfp32 · [2048, 128]
NVIDIA GeForce RTX 3090
13.3µs
#1 of 2
2025-04-30
Softmaxbf16 · [2048, 256]
NVIDIA GeForce RTX 3090
13.3µs
#1 of 2
2025-04-30
Softmaxbf16 · [2048, 1024]
NVIDIA GeForce RTX 3090
14.3µs
#2 of 2
2025-04-30
Show all 24 measurements ›
Softmaxfp32 · [2048, 256]
NVIDIA GeForce RTX 3090
19.2µs
#1 of 2
2025-04-30
Softmaxbf16 · [2048, 512]
NVIDIA GeForce RTX 3090
19.5µs
#1 of 2
2025-04-30
Softmaxfp32 · [2048, 1024]
NVIDIA GeForce RTX 3090
22.5µs
#2 of 2
2025-04-30
Softmaxfp32 · [2048, 512]
NVIDIA GeForce RTX 3090
38.9µs
#1 of 2
2025-04-30
Softmaxbf16 · [2048, 1024]
NVIDIA GeForce RTX 3090
42.0µs
#1 of 2
2025-04-30
Softmaxfp32 · [2048, 2048]
NVIDIA GeForce RTX 3090
57.6µs
#2 of 2
2025-04-30
Softmaxbf16 · [2048, 2048]
NVIDIA GeForce RTX 3090
59.4µs
#2 of 2
2025-04-30
Softmaxbf16 · [2048, 4096]
NVIDIA GeForce RTX 3090
67.6µs
#2 of 2
2025-04-30
Softmaxfp32 · [2048, 4096]
NVIDIA GeForce RTX 3090
83.2µs
#2 of 2
2025-04-30
Softmaxfp32 · [2048, 1024]
NVIDIA GeForce RTX 3090
84.0µs
#1 of 2
2025-04-30
Softmaxbf16 · [2048, 2048]
NVIDIA GeForce RTX 3090
126.0µs
#1 of 2
2025-04-30
Softmaxfp32 · [2048, 2048]
NVIDIA GeForce RTX 3090
183.3µs
#2 of 2
2025-04-30
Softmaxbf16 · [2048, 4096]
NVIDIA GeForce RTX 3090
197.6µs
#2 of 2
2025-04-30
Softmaxfp32 · [2048, 4096]
NVIDIA GeForce RTX 3090
330.8µs
#2 of 2
2025-04-30

Reported · How evidence levels are derived →

Source and license

sourceavailable
revision digestsha256:643178d96c20c26e7a3c3c424ad25f85170a76224689482539da1059ff9aacea
license declaredBSD-3-Clause
license concludedBSD-3-Clause
imported2026-08-18

Best evidence level for this revision: reported

JSON