Distributed Training
التدريب الموزَّع
أسلوب تدريب يُقسَّم فيه العمل الحوسبي على عدة وحدات معالجة (عادةً وحدات رسومية) عبر خوادم متعددة، بحيث تعمل جميعها بالتوازي لتقليص زمن التدريب مع الحفاظ على جودة النموذج.
A training approach where computational work is divided across multiple processing units (typically GPUs) across multiple servers, all working in parallel to reduce training time while maintaining model quality.
Also translated asالتدريب على عدة وحدات، التدريب المتوازي الموزَّع
First appears in this corpus in: Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour (2017)
Appears in these papers
- Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour2017in the sky ✦
- Learning Dexterous In-Hand Manipulation2019in the sky ✦
- Learning Dexterous In-Hand Manipulation2019in the sky ✦
- Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism2019in the sky ✦
- OPT: Open Pre-Trained Transformer Language Models2022in the sky ✦
- QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation2018in the sky ✦