YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
PyTorch mobile LIST_CONSTRUCT stack underflow PoC
This repository demonstrates a stack-underflow crash in PyTorch's lite-interpreter bytecode runtime.
What this PoC shows:
- a valid
.ptlmodel can be mutated by changing only one serializedLIST_CONSTRUCTarity operand - the original model still runs normally when
LIST_CONSTRUCT.Nstays at2 - changing
LIST_CONSTRUCT.Nfrom2to3makesforward()crash the officialtorch-2.13.0+cpuwheel withSIGSEGV - the malformed model reproduces through both
BytesIOand file-path loads
What this PoC does not show:
- a stable memory disclosure primitive
- code execution
- a claim about every TorchScript loading path
Verified runtime:
- official
torch-2.13.0+cpuwheel
Run:
python3 poc_mobile_list_construct_stack_underflow.py --list-n 2
python3 poc_mobile_list_construct_stack_underflow.py --list-n 3
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support