AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer

Greg Yang, Edward J Hu, Igor Babuschkin, Szymon Sidor, Xiaodong Liu, David Farhi, Nick Ryder, Jakub Pachocki, Weizhu Chen, Jianfeng Gao

NeurIPS 2021 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging · 11 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
transfer HPs from 0.25x proxy via muP instead of tuning the 1x model directly (equal compute)IWSLT14 Transformer · IWSLT14 De-Enmedian val BLEU: 35.00 -> 35.33Improved
use muP parametrization instead of SP when transferring HPs (ablation)IWSLT14 Transformer · IWSLT14 De-EnBLEU: training diverged -> 35.33Improved
muTransfer vs tuning 1x with same number of samplesIWSLT14 Transformer · IWSLT14 De-EnBLEUMixed
muTransfer from 15M proxyWMT14 Transformer 211M · WMT14 En-Demedian val BLEU: 25.69 -> 26.34Improved
muTransfer from BERT-prototypeBERT-large · BERT pretrainingtest loss: 1.731 -> 1.683Improved
muTransfer from BERT-prototypeBERT-base · BERT pretrainingtest loss: 1.995 -> 1.970Improved
muTransfer HPs on GPT-3 6.7BGPT-3 6.7B · HellaSwag Zero-Shotaccuracy: 66.7 -> 72.0Improved
muTransfer HPs on GPT-3 6.7BGPT-3 6.7B · LAMBADAaccuracy: Zero-shot 70.8 -> 73.5; Few-shot 77.1 -> 74.7Mixed
transfer best init std across depth2-layer Transformer muP · Wikitext-2loss optimumHurt

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