Xin Liu

DeepSeek-AI, Xiao Bi, Deli Chen et al. · microsoft-research, pku

The rapid development of open-source large language models (LLMs) has been truly remarkable. However, the scaling law described in previous literature presents varying conclusions, which casts a dark cloud over scaling LLMs. We delve into the study of scaling laws and present our distinctive findings that facilitate scaling of large scale models in two commonly used open-source configurations, 7B and 67B. Guided by the scaling laws, we introduce DeepSeek LLM, a project dedicated to advancing open-source language models with a long-term perspective. To support the pre-training phase, we have developed a dataset that currently consists of 2 trillion tokens and is continuously expanding. We further conduct supervised fine-tuning (SFT) and Direct Preference Optimization (DPO) on DeepSeek LLM Base models, resulting in the creation of DeepSeek Chat models. Our evaluation results demonstrate that DeepSeek LLM 67B surpasses LLaMA-2 70B on various benchmarks, particularly in the domains of code, mathematics, and reasoning. Furthermore, open-ended evaluations reveal that DeepSeek LLM 67B Chat exhibits superior performance compared to GPT-3.5.

Pengchao Han, Chao Huang, Geng Tian et al.

Split federated learning (SFL) is a recent distributed approach for collaborative model training among multiple clients. In SFL, a global model is typically split into two parts, where clients train one part in a parallel federated manner, and a main server trains the other. Despite the recent research on SFL algorithm development, the convergence analysis of SFL is missing in the literature, and this paper aims to fill this gap. The analysis of SFL can be more challenging than that of federated learning (FL), due to the potential dual-paced updates at the clients and the main server. We provide convergence analysis of SFL for strongly convex and general convex objectives on heterogeneous data. The convergence rates are $O(1/T)$ and $O(1/\sqrt[3]{T})$, respectively, where $T$ denotes the total number of rounds for SFL training. We further extend the analysis to non-convex objectives and the scenario where some clients may be unavailable during training. Experimental experiments validate our theoretical results and show that SFL outperforms FL and split learning (SL) when data is highly heterogeneous across a large number of clients.

Justin Dachille, Chao Huang, Xin Liu

Split Federated Learning (SFL) is a distributed machine learning paradigm that combines federated learning and split learning. In SFL, a neural network is partitioned at a cut layer, with the initial layers deployed on clients and remaining layers on a training server. There are two main variants of SFL: SFL-V1 where the training server maintains separate server-side models for each client, and SFL-V2 where the training server maintains a single shared model for all clients. While existing studies have focused on algorithm development for SFL, a comprehensive quantitative analysis of how the cut layer selection affects model performance remains unexplored. This paper addresses this gap by providing numerical and theoretical analysis of SFL performance and convergence relative to cut layer selection. We find that SFL-V1 is relatively invariant to the choice of cut layer, which is consistent with our theoretical results. Numerical experiments on four datasets and two neural networks show that the cut layer selection significantly affects the performance of SFL-V2. Moreover, SFL-V2 with an appropriate cut layer selection outperforms FedAvg on heterogeneous data.