[Feature Request] can we gather some math and code benchmark pipline to evaluate any training model which use our datagen pipline

[zjrwtx]

Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

after we training a model which use our datagen pipline ,we shoud have a lightbenchmark to evaluate the results, may be code and maths is the most important scope to evaulate

Solution

No response

Alternatives

No response

Additional context

No response

[hallerite]

Would GSM8K and MATH be okay as a start?