Skip to content

Latest commit

 

History

History
37 lines (26 loc) · 1.77 KB

README.md

File metadata and controls

37 lines (26 loc) · 1.77 KB
Raw

Code-Vectors-Artifact Dataset

If you haven't already run the artifact

The following commands will get you setup. WARNING: this artifact produces several gigabytes of output data!

To get setup with our original traces and two sets of learned vectors run the following:

# 1. Make sure we have an output directory to hold the data
mkdir -p ./output
# 2. Pulls the docker image down from docker hub
docker pull jjhenkel/code-vectors-artifact:dataset
# 3. Extracts the (heavily compressed) datasets to the ./output dir
docker run --rm -it -v `pwd`/output:/output jjhenkel/code-vectors-artifact:dataset

That's it, you're all set!

If you've already run the artifact

If you're reading this after running a container off of the jjhenkel/code-vectors-artifact:dataset docker image you should have a directory (./output if you followed the steps above) with the following files:

  • README.md: this readme.
  • vectors-vmin-0.txt: our vectors learned without a vocabulary minimum (vmin) threshold set.
  • vectors-vmin-1000.txt: our vectors learned with a more reasonable vocabulary minimum (vmin) threshold (to filter out rare words).
  • kernel-traces: our full set of abstracted symbolic traces extracted from the Linux kernel using our spec2image, c2ocaml, and lsee tools.
  • peek-traces.sh: a tool to run to view the first N traces from our trace corpus (streams and decompresses the gzipped traces and then runs head -n$1). Example: ./peek-traces.sh 100 (prints the first 100 traces from our corpus). NOTE: you need to run this tool in the ./output directory.
  • analogies.py: our full analogies benchmark and scaffolding code to run it on a given set of traces. (NOTE: requires Python and the Gensim library to be installed/working.)