CVE-2025-46153 — Inefficient CPU Computation in Pytorch

CWE-1176 — Inefficient CPU ComputationCWE-440 — Expected Behavior Violation6 documents6 sources
Severity
5.3MEDIUMNVD
EPSS
0.1%
top 78.73%
CISA KEV
Not in KEV
Exploit
No known exploits
Affected products
4
Linuxfoundation PytorchDebian PytorchMsrc Azl3 Pytorch 2.2.2-7 ON Azure Linux 3.0+1 more
Timeline
PublishedSep 25

Description

PyTorch before 3.7.0 has a bernoulli_p decompose function in decompositions.py even though it lacks full consistency with the eager CPU implementation, negatively affecting nn.Dropout1d, nn.Dropout2d, and nn.Dropout3d for fallback_random=True.

CVSS vector

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:NExploitability: 3.9 | Impact: 1.4

Affected Packages4 packages

▶NVDlinuxfoundation/pytorch2.6.0 — 2.7.0

Patches

Patch: github.com ↗

🔴Vulnerability Details

2
GHSA
GHSA-h36m-9qjr-pv93: PyTorch before 3↗2025-09-25
▶
OSV
CVE-2025-46153: PyTorch before 3↗2025-09-25
▶

📋Vendor Advisories

3
Red Hat
torch: PyTorch inconsistent results↗2025-09-25
▶
Microsoft
PyTorch before 3.7.0 has a bernoulli_p decompose function in decompositions.py even though it lacks full consistency with the eager CPU implementation, negatively affecting nn.Dropout1d, nn.Dropout2d,↗2025-09-09
▶
Debian
CVE-2025-46153: pytorch - PyTorch before 3.7.0 has a bernoulli_p decompose function in decompositions.py e...↗2025
▶