submission 918521
Purshotam Kumar · python · License unknown
Use it
Vendorable · source mirrored · license unknownView source →
No package. Vendor the mirrored source: 19 lines, June 9 Researcher Reciprocity License v1.0.
cholesky.py
curl "https://kernelindex.com/api/v1/implementations/kernelbot-cholesky-918521?include=source"interfacepython
Compatibility
measured onNVIDIA B200
declared hardwareNVIDIA B200
architecturessm_100
dtypesfp32
Benchmark evidence
1 measurement across 1 GPU, fastest first.
Operation / workload
Hardware
Latency
Rank
Observed
Reported · How evidence levels are derived →
Source and license
sourceavailable
revision digestsha256:27a2972ded106d27add4b2fdfc589a339045701c9a2ff1d70e75cd28474360a4
license declaredunknown
license concludedunknown
authorsPurshotam Kumar
imported2026-08-26
Kernel source
cholesky.py19 lines
import torch
from task import input_t, output_t
def custom_kernel(data: input_t) -> output_t:
batch, n, _ = data.shape
A = data.contiguous()
if n >= 8192 and batch == 1 and A.is_cuda:
torch.cuda.empty_cache()
L = torch.linalg.cholesky(A, upper=False)
diag = L.diagonal(dim1=-2, dim2=-1)
if torch.any(diag <= 0):
L = L.clone()
L.diagonal(dim1=-2, dim2=-1).copy_(torch.clamp(diag, min=1e-8))
return LSource code from GPU Mode and the KernelBot dataset · June 9 Researcher Reciprocity License v1.0
Best evidence level for this revision: reported
JSON