Meta-Eval/EvalHarness
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix NameError _fmt (module-level helper restored); aime/hmmt max_tokens 32768->8192 (non-stream path, gateway honors nothink)

Browse Source 
Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
sora 2026-09-11 07:20:02 +00:00
parent d64b236540
commit 63ffd1b613
5 changed files with 10 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
evalharness/data/datasets/aime24.py
View File

@ -13,7 +13,7 @@ from ..spec import DatasetSpec
name='aime24', name='aime24',
source='HuggingFaceH4/aime_2024', # https://huggingface.co/datasets/HuggingFaceH4/aime_2024 source='HuggingFaceH4/aime_2024', # https://huggingface.co/datasets/HuggingFaceH4/aime_2024
split='train', # the dataset ships a single split split='train', # the dataset ships a single split
gen_config={'temperature': 1.0, 'max_tokens': 32768}, gen_config={'temperature': 1.0, 'max_tokens': 8192},
prompt_style='aime_es', # es MathArena template: instruction-first, prompt_style='aime_es', # es MathArena template: instruction-first,
task_type='math', task_type='math',
tags=['math', 'competition'], tags=['math', 'competition'],

2
evalharness/data/datasets/aime25.py
View File

@ -10,7 +10,7 @@ from ..spec import DatasetSpec
name='aime25', name='aime25',
source='yentinglin/aime_2025', # https://huggingface.co/datasets/yentinglin/aime_2025 source='yentinglin/aime_2025', # https://huggingface.co/datasets/yentinglin/aime_2025
split='train', split='train',
gen_config={'temperature': 1.0, 'max_tokens': 32768}, gen_config={'temperature': 1.0, 'max_tokens': 8192},
prompt_style='aime_es', # es MathArena template: instruction-first, prompt_style='aime_es', # es MathArena template: instruction-first,
task_type='math', task_type='math',
tags=['math', 'competition'], tags=['math', 'competition'],

2
evalharness/data/datasets/aime26.py
View File

@ -10,7 +10,7 @@ from ..spec import DatasetSpec
name='aime26', name='aime26',
source='MathArena/aime_2026', # curated by MathArena (HuggingFace) source='MathArena/aime_2026', # curated by MathArena (HuggingFace)
split='train', # the dataset ships a single split split='train', # the dataset ships a single split
gen_config={'temperature': 1.0, 'max_tokens': 32768}, gen_config={'temperature': 1.0, 'max_tokens': 8192},
prompt_style='aime_es', # es MathArena template: instruction-first, prompt_style='aime_es', # es MathArena template: instruction-first,
task_type='math', task_type='math',
tags=['math', 'competition'], tags=['math', 'competition'],

2
evalharness/data/datasets/hmmt26.py
View File

@ -10,7 +10,7 @@ from ..spec import DatasetSpec
name='hmmt26', name='hmmt26',
source='MathArena/hmmt_feb_2026', # curated by MathArena (HuggingFace) source='MathArena/hmmt_feb_2026', # curated by MathArena (HuggingFace)
split='train', # the dataset ships a single split split='train', # the dataset ships a single split
gen_config={'temperature': 1.0, 'max_tokens': 32768}, gen_config={'temperature': 1.0, 'max_tokens': 8192},
prompt_style='imo_es', # es template: Problem: prefix + boxed suffix, prompt_style='imo_es', # es template: Problem: prefix + boxed suffix,
task_type='math', task_type='math',
tags=['math', 'competition'], tags=['math', 'competition'],

6
evalharness/progress/rich_terminal.py
View File

@ -13,6 +13,12 @@ from rich.progress import (
) )
def _fmt(sec):
sec = int(sec)
return f'{sec // 60}m{sec % 60:02d}s' if sec >= 60 else f'{sec}s'
class RichTerminalProgress: class RichTerminalProgress:
def __init__(self, console=None): def __init__(self, console=None):
# accept an EXTERNAL console: CLI phase messages and the live bar must # accept an EXTERNAL console: CLI phase messages and the live bar must