The paper introduces a hardware-software co‑design framework that compresses Mixture‑of‑Experts (MoE) model weights into low‑precision, hardware‑native sparse representations, enabling efficient execution on Sparse Tensor Cores (SpTCs). By relaxing discrete support selection through continuous reparameterization, the method jointly optimizes quantized weights and supports a router‑weighted reconstruction objective, achieving up to 4.35 percentage‑point gains in joint sparse‑quantization accuracy while retaining 96.09% of the original model’s performance. A custom grouped sparse GEMM kernel further boosts inference speed, outperforming NVIDIA’s baseline by up to 1.65× and reducing latency by up to 4.03× on B200 GPUs.
Semi-structured 2:4 sparsity is widely supported by modern accelerators, providing up to a 2x theoretical speedup. However, its strict 50% sparsity constraint often causes non-negligible accuracy degradation under post-training pruning.
arXiv:2608. 01536v1 Announce Type: cross Abstract: Large Language Models (LLMs) increasingly rely on sparsity to reduce inference cost, but most prior work targets a single sparsity source-either weight or activation-and optimizes for batched multi-user inference.
arXiv:2606. 10445v1 Announce Type: new Abstract: Semi-structured 2:4 sparsity is widely supported by modern accelerators, providing up to a 2x theoretical speedup.
Celty introduces a co-designed sparse format, GPU kernel, and SIMT microarchitecture to efficiently handle Sparse Matrix‑Sparse Vector (SpMSpV) workloads in large language model inference. Its Run‑Length Compressed CSC (RLC‑CSC) format allows vectorized loading of compressed weight columns and skips memory accesses by exploiting both weight pruning and activation sparsity. The Celty Sparse SIMT Core adds a pipelined RLC decoder that eliminates software index reconstruction and uses local registers for conflict‑free accumulation, achieving up to 5.3× speedup over cuBLAS at 70% dual‑sparsity.
arXiv:2608. 15602v1 Announce Type: cross Abstract: While binary quantization theoretically promises extreme compression and acceleration for Large Language Models (LLMs), existing research often overlooks the necessity of specialized hardware kernels, thus failing to unleash the full acceleration potential due to persistent reliance on expensive floating-point arithmetic or runtime dequantization overheads.
arXiv:2607. 08786v1 Announce Type: cross Abstract: With the growing deployment of large language models (LLMs), LLM inference cost has become a key challenge.
arXiv:2607. 25504v1 Announce Type: cross Abstract: Fine-grained weight pruning and activation sparsification have emerged as effective approaches for reducing the compute and memory cost of inference for Transformer models.
arXiv:2606. 01117v1 Announce Type: cross Abstract: Extreme multi-label classification (XMC) involves learning models over large output spaces with millions of labels, making the output layer a memory-compute bottleneck.
The paper presents a budget‑aware compression pipeline for deploying 70B‑parameter language models on a single NVIDIA GPU. It examines how pruning, quantization, and KV‑cache compression interact, showing that layer‑wise pruning improves weight quantization robustness and that KV‑cache sparsification complements INT8 KV quantization without harming decoding speed. Using these insights, the authors compressed a 70B model to ~33 GB, achieving ~57 tokens/s on 10k‑token prompts on an A40 while maintaining accuracy within 5% on standard benchmarks.
arXiv:2607. 21985v1 Announce Type: cross Abstract: The increasing deployment of large language models (LLMs) has magnified the computational and memory bottlenecks of autoregressive decoding, where low compute intensity and bandwidth-bound kernels dominate inference cost.
SparseDitto is an agentic sparse compilation framework that jointly synthesizes representation, execution schedule, and hardware mapping for sparse matrix computations on GPUs. It uses structural analysis, a learned template-ranking prior, and LLM-guided lowering to generate CUDA code, with target-GPU profiling refining the plan. The framework supports multiple operators such as SpMV, SpMM, and SpGEMM, adapts to different hardware, and achieves significant speedups over cuSPARSE, including up to 146.61× on certain matrices and 3.39× acceleration for full-batch GCN training.
arXiv:2609.38090v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) models are a compelling architecture for scaling model capacity, making them especially attractive for deployment on resource-...