fix(curriculum): Make boundaries in curriculum more sensible (#407)

* init

* fix tests

* unify codeio

* filtered for libraries not present in reasoning-gym

* fix more bounds

* puzzle24

* knight swap curriculum

* fix number sorting

* fix attributes

* add validation of config in creation of dataset

* dry run for instantiating and validating the datasets

* remove unused imports

* fix curriculum tests to reference newly updated attribute names
This commit is contained in:
Zafir Stojanovski 2025-04-04 20:24:14 +02:00 committed by GitHub
parent 7853263650
commit dced3bfc45
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
132 changed files with 1226 additions and 347 deletions

View file

@ -131,8 +131,8 @@ class CalendarArithmeticDataset(ProceduralDataset):
metadata["source_dataset"] = DATASET_NAME
metadata["source_index"] = idx
metadata["difficulty"] = {
"task_complexity": self.tasks.index(task),
"date_range": self.config.offset_upper_bound,
"tasks": self.config.tasks,
"offset_upper_bound": self.config.offset_upper_bound,
}
return {
"question": question,
@ -500,7 +500,7 @@ class CalendarArithmeticCurriculum(BaseCurriculum):
# Define attributes
self._define_attributes(
ScalarAttributeDefinition(
name="task_complexity",
name="tasks",
levels=[
["weekday_of_date"],
["weekday_of_date", "is_leap_year", "weekday_offset"],
@ -519,7 +519,7 @@ class CalendarArithmeticCurriculum(BaseCurriculum):
field_name="tasks",
),
ScalarAttributeDefinition(
name="date_range",
name="offset_upper_bound",
levels=[30, 100, 250, 365],
description="Maximum day range for offset and counting tasks",
field_name="offset_upper_bound",