Depeng Liang

Pengfei Zuo, Huimin Lin, Junbo Deng et al.

The rapid evolution of large language models (LLMs), driven by growing parameter scales, adoption of mixture-of-experts (MoE) architectures, and expanding context lengths, imposes unprecedented demands on AI infrastructure. Traditional AI clusters face limitations in compute intensity, memory bandwidth, inter-chip communication, and latency, compounded by variable workloads and strict service-level objectives. Addressing these issues requires fundamentally redesigned hardware-software integration. This paper introduces Huawei CloudMatrix, a next-generation AI datacenter architecture, realized in the production-grade CloudMatrix384 supernode. It integrates 384 Ascend 910 NPUs and 192 Kunpeng CPUs interconnected via an ultra-high-bandwidth Unified Bus (UB) network, enabling direct all-to-all communication and dynamic pooling of resources. These features optimize performance for communication-intensive operations, such as large-scale MoE expert parallelism and distributed key-value cache access. To fully leverage CloudMatrix384, we propose CloudMatrix-Infer, an advanced LLM serving solution incorporating three core innovations: a peer-to-peer serving architecture that independently scales prefill, decode, and caching; a large-scale expert parallelism strategy supporting EP320 via efficient UB-based token dispatch; and hardware-aware optimizations including specialized operators, microbatch-based pipelining, and INT8 quantization. Evaluation with the DeepSeek-R1 model shows CloudMatrix-Infer achieves state-of-the-art efficiency: prefill throughput of 6,688 tokens/s per NPU and decode throughput of 1,943 tokens/s per NPU (<50 ms TPOT). It effectively balances throughput and latency, sustaining 538 tokens/s per NPU even under stringent 15 ms latency constraints, while INT8 quantization maintains model accuracy across benchmarks.

Depeng Liang, Yongdong Zhang

Recently deeplearning models have been shown to be capable of making remarkable performance in sentences and documents classification tasks. In this work, we propose a novel framework called AC-BLSTM for modeling sentences and documents, which combines the asymmetric convolution neural network (ACNN) with the Bidirectional Long Short-Term Memory network (BLSTM). Experiment results demonstrate that our model achieves state-of-the-art results on five tasks, including sentiment analysis, question type classification, and subjectivity classification. In order to further improve the performance of AC-BLSTM, we propose a semi-supervised learning framework called G-AC-BLSTM for text classification by combining the generative model with AC-BLSTM.