CerebrasBlog11 MarAccelerating Large GPT Training with Sparse Pre-Training and Dense Fine-Tuning [Updated]