4 ms·
This is similar to why increasing the batch size during LLM training results in better performance: you force the optimizer to generalize to a larger set.
by reissbaker 2y ago
This is similar to why increasing the batch size during LLM training results in better performance: you force the optimizer to generalize to a larger set.