Skip to content

Conversation

@thomasRoglin
Copy link
Contributor

This PR aims to introduce a verification system inside the autohecbench.py allowing the benchmark results to be flagged as correct or not in the summary output.

In the subset.json, each benchmark entry now includes a verification section :

[
    verifcation_type,
    [param1, param2, param3 ... ]
]

For now, only 2 verification types are supported :

  • no_verification: means that no verification is implemented / configured
  • verification_token: Takes two parameters: [success_token, fail_token].
    The benchmark output is considered valid if it contains the success_token at least once and the fail_token does not appear.

In the autohecbench.py script, we introduce a new argument --verify that will activate the verification.
When enabled, the run() function checks the verification type of the benchmark and performs the related verification.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant