论文标题
部分可观测时空混沌系统的无模型预测
PARAGEN : A Parallel Generation Toolkit
论文作者
论文摘要
Paragen是一种基于Pytorch的NLP工具包，用于并行生成。 Paragen提供了13种类型的可自定义插件，可帮助用户快速尝试跨模型架构，优化和学习策略的新颖想法。我们实施各种功能，例如无限的数据加载和自动模型选择，以增强其工业用法。现在，Paragen被部署以支持BOCTEDANCE的各种研究和行业应用。 Paragen可在https://github.com/bytedance/paragen上找到。
PARAGEN is a PyTorch-based NLP toolkit for further development on parallel generation. PARAGEN provides thirteen types of customizable plugins, helping users to experiment quickly with novel ideas across model architectures, optimization, and learning strategies. We implement various features, such as unlimited data loading and automatic model selection, to enhance its industrial usage. ParaGen is now deployed to support various research and industry applications at ByteDance. PARAGEN is available at https://github.com/bytedance/ParaGen.
