Picture for Jiawen Niu

Jiawen Niu

Demystifying Workload Imbalances in Large Transformer Model Training over Variable-length Sequences

Add code
Dec 10, 2024
Figure 1 for Demystifying Workload Imbalances in Large Transformer Model Training over Variable-length Sequences
Figure 2 for Demystifying Workload Imbalances in Large Transformer Model Training over Variable-length Sequences
Figure 3 for Demystifying Workload Imbalances in Large Transformer Model Training over Variable-length Sequences
Figure 4 for Demystifying Workload Imbalances in Large Transformer Model Training over Variable-length Sequences
Viaarxiv icon