search

ASSERT-KTH / VRepair

open science repo of "Neural Transfer Learning for Repairing Security Vulnerabilities in C Code" https://arxiv.org/pdf/2104.08308
57 stars 18 forks source link

Update gensrctgt.py #2

Closed chenzimin closed 3 years ago

chenzimin commented 4 years ago

Add S2SV_ModEnd and after-context token for code addition.

chenzimin commented 4 years ago

@SteveKommrusch WDYT about making this change?

SteveKommrusch commented 4 years ago

Interesting to consider. It makes the output sequence longer but more regular. It's not obvious to me if it would be better or worse but we could try it. I would want to change chksrctgt.py too to double-check that the change is correct. I'd probably add it as an option so we can switch between the 2 possibilities. Given that we'll be talking a fair bit about the context idea, I think this could be a useful ablation study. If we do this, I think I should create 'data/MoreContext2,3,4,...' directories so that we still have the shorter context for comparison.