Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Restrict the format of acceptable followup completions #44

Merged
merged 5 commits into from
Oct 25, 2023
Merged

Restrict the format of acceptable followup completions #44

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
d9d17f8
Restrict the format of acceptable followup completions
steffencruz Oct 23, 2023
5c3a60f
Limit quaestion to 40 words
steffencruz Oct 24, 2023
92c7988
Black
steffencruz Oct 24, 2023
400deb7
Merge branch 'staging' into hotfix/short-followup
steffencruz Oct 24, 2023
1b27790
Merge branch 'staging' into hotfix/short-followup
p-ferreira Oct 25, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 16 additions & 0 deletions prompting/validators/forward.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -98,6 +98,22 @@ async def run_step(
timeout=timeout,
)

# Restrict the format of acceptable followup completions.
for response in responses:
# remove leading and trailing periods
completion = response.completion.strip(".")

if "followup" in name and len(completion) > 0:
if "?" in completion:
# take first question that is found and only use the sentence before the question mark
completion = completion.split("?")[0].split(".")[-1]
else:
# otherwise take the last sentence
completion = completion.split(".")[-1].split(".")[-1]

# take maximum of 40 words
response.completion = " ".join(completion.split(" ")[-40:]) + "?"

# Compute the rewards for the responses given the prompt.
rewards: torch.FloatTensor = torch.zeros(len(responses), dtype=torch.float32).to(
self.device
Expand Down