wassname/Open-Assistant
1
0
Fork 0
mirror of https://github.com/wassname/Open-Assistant.git synced 2026-06-29 16:30:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix typos (#1143)

Browse Source 
Found via `codespell -S .mypy_cache,yarn.lock,*.json,*.ipynb -L
rouge,nam,vie`
This commit is contained in:
Kian-Meng Ang
2023-02-06 03:18:03 +08:00
committed by GitHub
parent 364a4f5aa3
commit 1e321a6fca
32 changed files with 50 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scripts/postprocessing/rankings.py
+1 -1
View File
@@ -66,7 +66,7 @@ def get_winner(pairs):
def get_ranking(pairs):
"""
Abuses concordance property to get a (not necessarily unqiue) ranking.
Abuses concordance property to get a (not necessarily unique) ranking.
The lack of uniqueness is due to the potential existence of multiple
equally ranked winners. We have to pick one, which is where
the non-uniqueness comes from
scripts/postprocessing/scoring.py
+3 -3
View File
@@ -58,7 +58,7 @@ def score_update_votes(new_vote: int, consensus: npt.ArrayLike, voter_data: Vote
after that voter cast a vote on a question.
This function is only to be run when archiving a question
i.e. the question has had sufficiently many votes, or we cann't get more than "K" bits of information
i.e. the question has had sufficiently many votes, or we can't get more than "K" bits of information
The consensus is the array of all votes cast by all voters for that question
We then update the voter data using the new information
@@ -88,7 +88,7 @@ def score_update_prompts(consensus: npt.ArrayLike, voter_data: Voter) -> Voter:
This function returns the gain of points for a given prompt's votes
In contrast to the other score updating functions, we can run this online as new votes come in.
i.e. the question has had sufficiently many votes, or we cann't get more than "K" bits of information.
i.e. the question has had sufficiently many votes, or we can't get more than "K" bits of information.
Parameters:
@@ -122,7 +122,7 @@ def score_update_ranking(user_ranking: npt.ArrayLike, consensus_ranking: npt.Arr
This function returns the gain of points for a given ranking's votes
This function is only to be run when archiving a question
i.e. the question has had sufficiently many votes, or we cann't get more than "K" bits of information
i.e. the question has had sufficiently many votes, or we can't get more than "K" bits of information
we use the bubble-sort distance (or "kendall-tau" distance) to compare the two rankings
we use this over spearman correlation since:
scripts/postprocessing/task_schedule.py
+1 -1
View File
@@ -56,7 +56,7 @@ def next_answer_task(possible_prompts, answers_per_prompt):
This helps to not have too much close-to-finished prompts in the active set.
Parameters:
possible_prompts (dict[prompt_id, num_answers]): a dictonary containing all open prompts and the number of answers these prompts currently have.
possible_prompts (dict[prompt_id, num_answers]): a dictionary containing all open prompts and the number of answers these prompts currently have.
answers_per_prompt (int): number of answers we per prompt to target
Returns:
prompt_id (int): the prompt_id corresponding to the next prompt that should get a new answer