Skip to content

feat: supported eval investigation and ran attacks in isolation#115

Draft
sdhossain wants to merge 2 commits into
mainfrom
sh/eval_investigation
Draft

feat: supported eval investigation and ran attacks in isolation#115
sdhossain wants to merge 2 commits into
mainfrom
sh/eval_investigation

Conversation

@sdhossain
Copy link
Copy Markdown
Collaborator

Changes

Summarize the changes in this PR and describe the context or motivation for
them.

Add a title, prepending the tag [attack], [defense], [evaluation], or [infra] if
appropriate.

Testing

Describe how you tested the changes in this PR. E.g., added tests, or ran
command foo and checked the results looked good.


if EvalName.MT_BENCH in self.attack_config.evals:
results = pl.concat([results, self.evaluate_mt_bench()])

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

can we throw an error if results is empty?

Otherwise if we add an eval and forget to add it here to evaluate(), it silently fails and we just don't get the evaluation results. It'd be nice for things to fail earlier.

Or another idea is to make use of the evaluation registry here so that we don't have to manually add each eval here in evaluate()

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.

2 participants