CVE-2025-46153
MEDIUM5.3EPSS 0.10%Published: 9/25/2025Modified: 5/20/2026
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.
Affected packages (3)
- Bitnami/pytorch>= 2.6.0, < 2.7.0
- Debian/pytorchfrom 0
- PyPI/torch>= 2.6.0, < 2.7.0
CVSS scores
| Source | Version | Severity | Vector |
|---|---|---|---|
| osv | CVSS 3.1 | MEDIUM5.3 | CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N |
References (7)
- ADVISORYhttps://gist.github.com/shaoyuyoung/4bcefba4004f8271e64b5185c95a248a
- ADVISORYhttps://gist.github.com/shaoyuyoung/e636f2e7a306105b7e96809e2b85c28a
- ADVISORYhttps://security-tracker.debian.org/tracker/CVE-2025-46153
- PATCHhttps://github.com/pytorch/pytorch/pull/143460
- REPORThttps://github.com/pytorch/pytorch/issues/142853
- WEBhttps://github.com/pytorch/pytorch/compare/v2.6.0...v2.7.0
- WEBhttps://nvd.nist.gov/vuln/detail/CVE-2025-46153