medium10 pts · first answer countsgpu-computingscaling-laws

Compute-optimal data size

The Chinchilla scaling study suggests a compute-optimal number of training tokens for a model of N parameters. Roughly how many tokens for a 7B model?

A.About 7 billion tokensB.About 7 trillion tokensC.About 1.4 billion tokensD.About 140 billion tokens

Discussion 0

Join free to post in the discussion.

Compute-optimal data size

Discussion 0

Related questions