AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Pipe-SGD: A Decentralized Pipelined SGD Framework for Distributed Deep Net Training

Youjie Li, Mingchao Yu, Songze Li, Salman Avestimehr, Nam Sung Kim, Alexander Schwing

NeurIPS 2018 · Architectures and efficiency › RNNs and normalization · approach: Distributed LLM inference systems · 64 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use decentralized pipelined SGD with one-iteration-delayed updates instead of D-SyncDistributed deep-net training · Four benchmarkswall-clock speedup: 2.0~3.2×Improved
Use decentralized pipelined SGD with one-iteration-delayed updates instead of PS-SyncDistributed deep-net training · Four benchmarkswall-clock speedup: 4.0~5.4×Improved
Replace PS-Sync with D-SyncDistributed training · CIFAR100-Convexconvergence speed: 40%Improved
Replace D-Sync with Pipe-SGDDistributed training · CIFAR100-Convexconvergence speed: 37%Improved
Add pipelining to D-Sync with truncationCIFAR100-Convex · CIFAR100-Convexconvergence speed: 46%Improved
Use decentralized balanced communication instead of a centralized parameter serverDistributed training · Four-node GPU clusteruncompressed communication time: 50%Improved
Add 16-bit truncation to Pipe-SGDPipe-SGD · AlexNettop-1 accuracy: -0.005Hurt
Add lossy compression to Pipe-SGDPipe-SGD · Benchmarks other than AlexNettop-1 accuracyImproved
Increase iteration dependency beyond K = 2 (ablation)Pipe-SGD · Distributed trainingwall-clock training timeHurt
Use sequential rather than pipelined gradient communication (ablation)Pipe-SGD when communication-bound · Distributed trainingwall-clock training timeImproved
Use complex compression instead of lightweight truncation or quantization (ablation)AllReduce · Pipe-SGD compression benchmarkscompression overhead relative to uncompressed communication: 1.6~2.3× the uncompressed communication time; 25.6~36.8× the compressed communication timeHurt
Use warm-up before enabling pipelining (ablation)Pipe-SGD on AlexNet and ResNet · Large non-convex modelsfinal accuracyImproved

Later papers that compared with this method

1 later accepted paper reports PipeSGD as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy