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

Improve reproducibility in utils_math.py #46

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Improve reproducibility in utils_math.py #46

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
14 changes: 7 additions & 7 deletions utils_math.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,12 +3,10 @@
import math
import inspect

from tqdm import tqdm
from tqdm.auto import trange
from sklearn.linear_model import LogisticRegression
from collections import defaultdict

np.random.seed(42)

STYLE_CONTROL_ELEMENTS = [
"sum_assistant_a_tokens",
"header_count_a",
Expand Down Expand Up @@ -74,8 +72,8 @@ def get_bootstrap_result(battles, func_compute_elo, num_round, baseline_model="g
kwargs = {}
if baseline_model in inspect.signature(func_compute_elo).parameters:
kwargs[baseline_model] = baseline_model
for _ in tqdm(range(num_round), desc="bootstrap"):
rows.append(func_compute_elo(battles.sample(frac=1.0, replace=True), **kwargs))
for round in trange(num_round, desc="bootstrap"):
rows.append(func_compute_elo(battles.sample(frac=1.0, replace=True, random_state=round), **kwargs))
df = pd.DataFrame(rows)
return df[df.median().sort_values(ascending=False).index]

Expand Down Expand Up @@ -206,9 +204,11 @@ def get_bootstrap_result_style_control(X, Y, battles, models, func_compute_elo,
X.shape[0] / 2
) # Since we duplicate the battles when constructing X and Y, we don't want to sample the duplicates

rng = np.random.default_rng(42)

battles_tie_idx = (battles["winner"] == "tie") | (battles["winner"] == "tie (bothbad)")
for _ in tqdm(range(num_round), desc="bootstrap"):
indices = np.random.choice(list(range(k)), size=(k), replace=True)
for _ in trange(num_round, desc="bootstrap"):
indices = rng.choice(list(range(k)), size=(k), replace=True)

index2tie = np.zeros(k, dtype=bool)
index2tie[battles_tie_idx] = True
Expand Down