CVE-2025-46153: Medium severity PyTorch PyTorch vulnerability
PyTorch before 3.7.0 has a bernoullip 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 fallbackrandom=True.
Affected Software
Remediation
Patch Available
Event History
Frequently Asked Questions
What is the severity of CVE-2025-46153?
CVE-2025-46153 has been classified as a moderate severity vulnerability that affects the functionality of dropout layers in PyTorch.
How do I fix CVE-2025-46153?
To fix CVE-2025-46153, update PyTorch to version 3.7.0 or later which addresses the inconsistency issue.
What software is affected by CVE-2025-46153?
CVE-2025-46153 affects versions of PyTorch prior to 3.7.0.
What problems does CVE-2025-46153 cause in PyTorch?
CVE-2025-46153 negatively affects nn.Dropout1d, nn.Dropout2d, and nn.Dropout3d leading to potential inconsistencies in random number generation during dropout.
Is there a workaround for CVE-2025-46153 if I cannot upgrade?
A temporary workaround for CVE-2025-46153 involves avoiding the use of dropout layers with fallback_random set to true until an upgrade can be performed.