Disclosure summary
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.
Source-reported weakness categories
CWE-1176
Source-specific records & product guidance
Sources retain their own attribution and scoring. Follow the original record to confirm affected versions, fixed releases, and configuration conditions.
Microsoft Security Updates (CVRF) · 2025-Sep
Open original source · Updated Aug 06, 2026
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.
Maximum of vendor-reported product scores; products and fixed builds are associations, not a universal affected-version statement.
Microsoft maximum product score: 5.3
| Vendor | Product / association | Version / bounds |
|---|---|---|
| Microsoft update guide | cbl2 pytorch 2.0.0-9 on CBL Mariner 2.0 | (MSRC status code 3) |
Original records & references
- NIST NVD record
- CVE Program record
- msrc.microsoft.com — Vendor advisory
PUBLISHED 2025-10-01T21:04:57-04:00
MODIFIED 2026-08-06T20:25:04-04:00
INGESTED 2026-10-08T12:45:41-04:00