{"action":"Upgrade pytorch to 2.12.0+dfsg2-1 or later.","affected":[{"fixed":"2.12.0+dfsg2-1","introduced":"0"}],"component":"pytorch","cve_ids":[],"ecosystem":"Debian","ecosystem_qualifier":"Debian:14","exploited":false,"fixed_versions":["2.12.0+dfsg2-1"],"id":"tcs_01M5NNXG45E4MY8M9V4SSZ35MA","modified_at":"2026-09-01T20:05:52.784247218Z","published_at":"2025-09-25T15:16:12.603Z","references":["https://osv.dev/vulnerability/DEBIAN-CVE-2025-46153","https://security-tracker.debian.org/tracker/CVE-2025-46153"],"severity":{"label":"medium","score":5.3},"sources":[{"source":"osv","source_id":"DEBIAN-CVE-2025-46153#Debian:14/pytorch","url":"https://osv.dev/vulnerability/DEBIAN-CVE-2025-46153"}],"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,"}