Skip to content

adding GEM/webnlg prompts #743

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.

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

Merged
merged 12 commits into from
May 8, 2022

Conversation

jordiclive
Copy link
Contributor

GEM/webnlg houses all constituents, and challenge sets. Also added Russian prompts with minimal changes, as input triple set is still english.

@jzf2101 jzf2101 self-requested a review April 26, 2022 22:18
@jzf2101 jzf2101 self-assigned this Apr 26, 2022
@jzf2101 jzf2101 changed the base branch from main to eval-hackathon April 26, 2022 23:26
Adding multi-target + PALM prompt
adding PALM and multi-target
@jordiclive jordiclive requested a review from awebson April 29, 2022 12:30
@jordiclive
Copy link
Contributor Author

I've added Multi-refs

@jon-tow
Copy link

jon-tow commented May 3, 2022

Hey @jordiclive ! When I run this I get a List[str] as opposed to the expected Tuple[str, List[str] return (see #747).

Example: For GEM/web_nlg ru

import datasets
from promptsource.templates import DatasetTemplates

template = DatasetTemplates("GEM/web_nlg", "ru")
data = datasets.load_dataset("GEM/web_nlg", "ru")
example = data['validation'][0]

prompt = template['PALM_prompt'].apply(example)
print(prompt)
[
'I will verbalize an abstract representation of a sentence in natural language. To do so, I will first show the representation and then the natural language. The text needs to include all of the information in the representation and it will be in Russian.\n\nPunjab,_Pakistan | leaderTitle | Provincial_Assembly_of_the_Punjab',
'Пенджаб, Пакистан, возглавляется Провинциальной ассамблеей Пенджаба.',
'Пенджаб, Пакистан возглавляется Провинциальной ассамблеей Пенджаба.'
]

@jordiclive
Copy link
Contributor Author

@jon-tow, I think you were not working with the newest eval-hackathon which has the multi-ref changes. I have now merged that into this .

@sebastianGehrmann
Copy link

Prompts look good to me. Do we have someone who could translate some of the Russian prompts so we don't only rely on English?

@jordiclive jordiclive merged commit 4927189 into bigscience-workshop:eval-hackathon May 8, 2022
Comment on lines +18 to +33
3e41305c-5461-4cf3-853d-8a6fb5747623: !Template
answer_choices: null
id: 3e41305c-5461-4cf3-853d-8a6fb5747623
jinja: "I will verbalize an abstract representation of a sentence in natural language.\
\ To do so, I will first show the representation and then the natural language.\
\ The text needs to include all of the information in the representation.\n\n\
{{input | join(\", \")}} {% for i in references %}\n ||| {{ i }} \n{% endfor\
\ %}"
metadata: !TemplateMetadata
choices_in_prompt: false
metrics:
- BLEU
- ROUGE
original_task: true
name: PALM_prompt
reference: used-in-PALM

Choose a reason for hiding this comment

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

how do we know this prompt was used in palm?
all i could find in their paper is the below ..
Screenshot 2022-08-15 at 22 11 45

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I believe @sebastianGehrmann shared some NLG prompts for PALM.

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.

5 participants