paper-with-me

홈 › Papers

PanGu-Σ: Towards Trillion Parameter Language Model with Sparse Heterogeneous Computing

2023-03-20 · Xiaozhe Ren, Pingyi Zhou, Xinfan Meng, Xinjing Huang, Yadao Wang, Weichao Wang, Pengfei Li, Xiaoda Zhang, Alexander Podolskiy, Grigory Arshinov, Andrey Bout, Irina Piontkovskaya, Jiansheng Wei, Xin Jiang, Teng Su, Qun Liu, Jun Yao

The scaling of large language models has greatly improved natural language understanding, generation, and reasoning. In this work, we develop a system that trained a trillion-parameter language model on a cluster of Ascend 910 AI processors and MindSpore framework, and present the language model with 1.085T parameters named PanGu-{\Sigma}. With parameter inherent from PanGu-{\alpha}, we extend the dense Transformer model to sparse one with Random Routed Experts (RRE), and efficiently train the model over 329B tokens by using Expert Computation and Storage Separation(ECSS). This resulted in a 6.3x increase in training throughput through heterogeneous computing. Our experimental findings show that PanGu-{\Sigma} provides state-of-the-art performance in zero-shot learning of various Chinese NLP downstream tasks. Moreover, it demonstrates strong abilities when fine-tuned in application data of open-domain dialogue, question answering, machine translation and code generation.

📄 PDF Abstract BibTeX arXiv:2303.10845

Code (0)

등록된 구현이 없습니다.

Tasks

Code GenerationLanguage ModelingLanguage ModellingMachine TranslationNatural Language UnderstandingQuestion AnsweringZero-Shot Learning

Methods 이 논문이 사용한 방법론

Multi-Head Attention 설명 없음
Attention 설명 없음
Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.
Dense Connections Dense Connections, or Fully Connected Connections, are a type of layer in a deep neural network that use a linear operation where every input is connected to every output…
Position-Wise Feed-Forward Layer 설명 없음
Adam 설명 없음
Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…
Label Smoothing Label Smoothing is a regularization technique that introduces noise for the labels. This accounts for the fact that datasets may have mistakes in them, so maximizing the…

Similar Papers 제목 키워드 기반

Pangu Ultra MoE: How to Train Your Big MoE on Ascend NPUs

2025-05-07 · Yehui Tang, Yichun Yin, Yaoyuan Wang, Hang Zhou 외

Sparse large language models (LLMs) with Mixture of Experts (MoE) and close to a trillion parameters are dominating the realm of most capable language models. However, the massive model scale poses significant challenges…

Mixture-of-Experts

Pangu Ultra: Pushing the Limits of Dense Large Language Models on Ascend NPUs

2025-04-10 · Yichun Yin, Wenyong Huang, Kaikai Song, Yehui Tang 외

We present Pangu Ultra, a Large Language Model (LLM) with 135 billion parameters and dense Transformer modules trained on Ascend Neural Processing Units (NPUs). Although the field of LLM has been witnessing unprecedented…

Large Language Model

Trillion Parameter AI Serving Infrastructure for Scientific Discovery: A Survey and Vision

2024-02-05 · Nathaniel Hudson, J. Gregory Pauloski, Matt Baughman, Alok Kamatar 외

Deep learning methods are transforming research, enabling new techniques, and ultimately leading to new discoveries. As the demand for more capable AI models continues to grow, we are now entering an era of Trillion Para…

scientific discovery

Every Activation Boosted: Scaling General Reasoner to 1 Trillion Open Language Foundation

2025-10-25 · Ling Team, Ang Li, Ben Liu, Binbin Hu 외 arxiv

We introduce Ling 2.0, a series reasoning-oriented language foundation built upon the principle that every activation boosts reasoning capability. Designed to scale from tens of billions to one trillion parameters under …

Computational Efficiency

PanGu-$α$: Large-scale Autoregressive Pretrained Chinese Language Models with Auto-parallel Computation

2021-04-26 · Wei Zeng, Xiaozhe Ren, Teng Su, Hui Wang 외

Large-scale Pretrained Language Models (PLMs) have become the new paradigm for Natural Language Processing (NLP). PLMs with hundreds of billions parameters such as GPT-3 have demonstrated strong performances on natural l…

Cloze (multi-choices) (Few-Shot)Cloze (multi-choices) (One-Shot)Cloze (multi-choices) (Zero-Shot)Common Sense Reasoning (Few-Shot)+17