Highly Available Data Parallel ML training on Mesh Networks

6 Nov 2020 · Sameer Kumar, Norm Jouppi ·

Data parallel ML models can take several days or weeks to train on several accelerators. The long duration of training relies on the cluster of resources to be available for the job to keep running for the entire duration. On a mesh network this is challenging because failures will create holes in the mesh. Packets must be routed around the failed chips for full connectivity. In this paper, we present techniques to route gradient summation allreduce traffic around failed chips on 2-D meshes. We evaluate performance of our fault tolerant allreduce techniques via the MLPerf-v0.7 ResNet-50 and BERT benchmarks. Performance results show minimal impact to training throughput on 512 and 1024 TPU-v3 chips.

PDF Abstract

Code

Add Remove Mark official

No code implementations yet. Submit your code now

Tasks

Add Remove

Datasets

Add Datasets introduced or used in this paper

Results from the Paper

Edit

Submit results from this paper to get state-of-the-art GitHub badges and help the community compare results to other papers.

Methods

Add Remove

Adam • Attention Dropout • BERT • Dense Connections • Dropout • GELU • Layer Normalization • Linear Layer • Linear Warmup With Linear Decay • Multi-Head Attention • Residual Connection • Scaled Dot-Product Attention • Softmax • Weight Decay • WordPiece

Edit Social Preview

Highly Available Data Parallel ML training on Mesh Networks

Code Edit Add Remove Mark official

Tasks Edit Add Remove

Datasets Edit

Results from the Paper Edit

Methods Edit Add Remove

Code

Add Remove Mark official

Tasks

Add Remove

Datasets

Results from the Paper

Edit

Methods

Add Remove