Score:1

What are the computational limitations of ZkProofs using ZoKrates?

ng flag

I'm trying to create a zk-proof of a neural network in https://github.com/berendjan/zk-neural-network using ZoKrates and PyTorch. The steps to reproduce are in the README.md

However, when computing the witness I run into errors

time ./run.sh 
Compiling network.zok

Compiled code written to 'out'
Number of constraints: 12697691
Performing setup...
Verification key written to 'verification.key'
Proving key written to 'proving.key'
Setup completed
Computing witness...
Execution failed: Sum check failed
The default ZoKrates interpreter should not yield this error. Please open an issue.

real    62m19.883s
user    50m2.462s
sys 16m59.227s

Am I hitting some limitations of zk-proofs with 12 million constraints? Am I hitting some ZoKrates limitations with 12 million constraints?

I want to build a verifier for proofs of a neural network computation, hoping that the computation will be much lower than otherwise needed when running the network directly. I'm using the parameters of the neural network model as constant fixed precision integers in the ZoKrates scripts. Does this work as I want too?

I sit in a Tesla and translated this thread with Ai:

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.