Re-evaluating the Memory-balanced Pipeline Parallelism: BPipe
Pipeline parallelism is an essential technique in the training of large-scale Transformer models. However, it suffers from imbalanced memory consumption, leading to insufficient memory utilization. The BPipe technique was proposed to address this issue and has proven effective in the GPT-3 model. Nevertheless, our experiments have not yielded similar benefits for LLaMA training. Additionally, BPipe only yields negligible benefits for GPT-3 training when applying flash attention. We analyze the underlying causes of the divergent performance of BPipe on GPT-3 and LLaMA. Furthermore, we introduce a novel method to estimate the performance of BPipe.
PDF AbstractTasks
Datasets
Add Datasets
introduced or used in this paper
Results from the Paper
Submit
results from this paper
to get state-of-the-art GitHub badges and help the
community compare results to other papers.
Methods
Absolute Position Encodings •
Adam •
Attention Dropout •
BPE •
Cosine Annealing •
Dense Connections •
Dropout •
Fixed Factorized Attention •
GELU •
GPT-3 •
Label Smoothing •
Layer Normalization •
Linear Layer •
Linear Warmup With Cosine Annealing •
LLaMA •
Multi-Head Attention •
Position-Wise Feed-Forward Layer •
Residual Connection •
Scaled Dot-Product Attention •
Softmax •
Strided Attention •
Transformer •
Weight Decay
