← All threads

Optimizer Scaling

Investigates how the choice of optimizer for pretraining interacts with the amount of training data to affect downstream fine-tuning performance.

1 papers

Where this stands

The written synthesis of this thread is for subscribers. Subscribe.