InternLM/InternBootcamp
1
0
Fork 0
mirror of https://github.com/InternLM/InternBootcamp.git synced 2026-04-23 16:55:02 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix bugs for symbolic regression bootcamp

Browse source
This commit is contained in:
lipeiji 2025-06-16 17:31:52 +08:00
parent 33101ef068
commit 5eb513f014
11 changed files with 148 additions and 64 deletions
Download patch file Download diff file Expand all files Collapse all files

9
internbootcamp/bootcamp/base.py
View file

@ -47,7 +47,7 @@ class Basebootcamp:
@classmethod
def verify_score(cls, model_output, identity: dict, format_score=0, short_penalty=False, short_threshold=256, ans_threshold=128, format_penalty=False) -> float:
def verify_score(cls, model_output, identity: dict, format_score=0, short_penalty=False, short_threshold=256, think_threshold=128, ans_threshold=128, format_penalty=False) -> float:
"""
Verify the output against the ground truth.
@ -83,10 +83,11 @@ class Basebootcamp:
pass
ans_output = model_output.rsplit("</think>", 1)[1] if "</think>" in model_output else ""
if (short_penalty and len(model_output) < short_threshold) or (short_penalty and len(ans_output) < ans_threshold):
think_length = len(model_output) - len(ans_output)
score = max(0, score) # Ensure score is non-negative
if (short_penalty and len(model_output) < short_threshold) or (short_penalty and len(ans_output) < ans_threshold) or (short_penalty and think_length < think_threshold):
# if the output is too short, consider it incorrect
return min(score * len(model_output) / short_threshold, score * len(ans_output) / ans_threshold)
return min(score * len(model_output) / short_threshold, score * len(ans_output) / ans_threshold, score * think_length / think_threshold)
# This for training Debug
if random.randint(1,1024) == 1: