Related papers: AMLA: MUL by ADD in FlashAttention Rescaling

Efficient inference of Multi-Head Latent Attention (MLA) is challenged by deploying the DeepSeek-R1 671B model on a single Multi-GPU server. This paper introduces FlashMLA-ETAP, a novel framework that enhances MLA inference for the…

Transformers and large language models (LLMs) have revolutionized machine learning, with attention mechanisms at the core of their success. As the landscape of attention variants expands, so too do the challenges of optimizing their…

While Transformer self-attention offers strong parallelism, the Key-Value (KV) cache grows linearly with sequence length and becomes a bottleneck for inference efficiency. Multi-head latent attention was recently developed to compress the…

Long-context inference in large language models is bottlenecked by Key--Value (KV) cache loading during the decoding stage, where the sequential nature of generation requires repeatedly transferring the KV cache from off-chip High-Bandwidth…

Multi-head Latent Attention (MLA) is an innovative architecture proposed by DeepSeek, designed to ensure efficient and economical inference by significantly compressing the Key-Value (KV) cache into a latent vector. Compared to MLA,…

Large language model (LLM) inference demands significant amount of computation and memory, especially in the key attention mechanism. While techniques, such as quantization and acceleration algorithms, like FlashAttention, have improved…

Long-rollout causal video diffusion has converged on a fixed-size sliding-window KV cache, with recent progress innovating within this layout by changing which tokens occupy the window or how their positions are encoded. The per-head KV…

Attention, as a core layer of the ubiquitous Transformer architecture, is the bottleneck for large language models and long-context applications. While FlashAttention-3 optimized attention for Hopper GPUs through asynchronous execution and…

All current LLM serving systems place the GPU at the center, from production-level attention-FFN disaggregation to NVIDIA's Rubin GPU-LPU heterogeneous platform. Even academic PIM/PNM proposals still treat the GPU as the central hub for…

Scaling depth is a key driver for large language models (LLMs). Yet, as LLMs become deeper, they often suffer from signal degradation: informative features formed in shallow layers are gradually diluted by repeated residual updates, making…

We present the first comprehensive study of latent multi-head attention (MLA) for small language models, revealing interesting efficiency-quality trade-offs. Training 30M-parameter GPT models on 100,000 synthetic stories, we benchmark three…

Transformer-based large language models (LLMs) have demonstrated remarkable performance across a wide range of real-world tasks, but their inference cost remains prohibitively high due to the quadratic complexity of attention and the memory…

The rise of disaggregated AI GPUs has exposed a critical bottleneck in large-scale attention workloads: non-uniform memory access (NUMA). As multi-chiplet designs become the norm for scaling compute capabilities, memory latency and…

The evolution of large language models (LLMs) towards applications with ultra-long contexts faces challenges posed by the high computational and memory costs of the Transformer architecture. While existing sparse and linear attention…

Multi-headed Attention's (MHA) quadratic compute and linearly growing KV-cache make long-context transformers expensive to train and serve. Prior works such as Grouped Query Attention (GQA) and Multi-Latent Attention (MLA) shrink the cache,…

Mixture of Block Attention (MoBA) (Lu et al., 2025) is a promising building block for efficiently processing long contexts in LLMs by enabling queries to sparsely attend to a small subset of key-value blocks, drastically reducing…

The advent of foundation models have revolutionized various fields, enabling unprecedented task accuracy and flexibility in computational linguistics, computer vision and other domains. Attention mechanism has become an essential component…

The Transformer architecture has significantly advanced natural language processing (NLP) and has been foundational in developing large language models (LLMs) such as LLaMA and OPT, which have come to dominate a broad range of NLP tasks.…

Transformers have significantly advanced AI and machine learning through their powerful attention mechanism. However, computing attention on long sequences can become a computational bottleneck. FlashAttention mitigates this by fusing the…

In this work, we study how multi-head latent attention (MLA), a popular strategy for compressing key/value memory, affects a transformer's internal capacity during pretraining. Using a lightweight suite of Marchenko-Pastur (MP) diagnostics,…