AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Sub-Linear Memory: How to Make Performers SLiM

Valerii Likhosherstov, Krzysztof Marcin Choromanski, Jared Quincy Davis, Xingyou Song, Adrian Weller

NeurIPS 2021 · Architectures and efficiency › Attention and transformers · approach: Sparse efficient transformers for long context · 5 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
reduce chunk size C from full to 64 (memory-efficient gradient)Performer with memory-efficient gradient · Copying task L=512peak GPU memory (GB): 0.0449 -> 0.0374Improved
reduce chunk size C from full to 64Performer with memory-efficient gradient · Copying task L=512time per iteration (sec): 0.0474 -> 0.1228Hurt
reduce chunk size C from full to 256Performer with memory-efficient gradient · Penn Treebank L=1024peak GPU memory (GB): 0.300 -> 0.231Improved
reduce chunk size C from full to 256Performer with memory-efficient gradient · Penn Treebank L=1024time per iteration (sec): 0.1377 -> 0.3060Hurt
reduce chunk size C from full to 1366Performer with memory-efficient gradient · Enwik8 L=4096peak GPU memory (GB): 1.513 -> 0.909Improved
reduce chunk size C from full to 1366Performer with memory-efficient gradient · Enwik8 L=4096time per iteration (sec): 0.4598 -> 0.8654Hurt
train with memory-efficient algorithm instead of full gradient, or fine-tune with itPerformer with memory-efficient gradient · Copying, PTB, Enwik8learning curvesNo clear effect
decrease C in gradient computationPerformer with memory-efficient gradient · random init modelsrelative gradient discrepancyHurt
use iterative instead of prefix-sum computationPerformer with memory-efficient gradient · Config II-IVtime and memoryMixed

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