arXiv:2610. 23796v1 Announce Type: new Abstract: Transformer prefill reuses weights across tokens and is dominated by matrix multiplication and activation data movement; single-token decode has limited reuse and is dominated by weight and KV-cache reads.
arXiv:2606. 14598v1 Announce Type: new Abstract: Post-training INT8 (W8A8) quantization of diffusion transformers is widely deployed as a speed optimization, yet on consumer Ampere GPUs it is frequently slower than the FP8 and NF4 alternatives it is meant to beat.
The paper introduces a multi‑shell decoder for Leech‑lattice vector quantization, achieving the best reported 2‑bit quality under its evaluation protocol. It presents a GPU‑friendly layout that fuses dequantization with matrix‑vector multiplication, demonstrating significant speed and memory advantages over traditional one‑hot masks and other 4‑bit methods. Experiments show the new kernel outperforms baseline approaches across multiple model sizes, with measurable gains in throughput and reduced byte traffic.
The paper shows that query‑based segmentation decoders on edge GPUs are limited by memory bandwidth rather than FLOPs. Profiling reveals that non‑tensor‑core operations consume a large fraction of decoder time and bytes, saturating the DRAM bus. CompactInst mitigates this by using parameter‑free spatial pooling and a lightweight refinement head, achieving significant speedups on Jetson Orin Nano while maintaining accuracy.
arXiv:2606. 08761v1 Announce Type: cross Abstract: W4A4 quantization promises full utilization of INT4 Tensor Cores, yet group dequantization overhead on CUDA Cores has driven existing systems to mixed-precision fallbacks.
arXiv:2607. 14568v1 Announce Type: cross Abstract: A companion study ran a 35B mixture-of-experts model on a 2011 NVIDIA Tesla C2075 (Fermi, sm_20, 6GB) as a GPU-prefill/CPU-decode hybrid, because the 4-bit model did not fit in device memory (arXiv:2606.
The paper introduces format‑aware fusion, a method that co‑designs quantization producers with their scale domains and consumer layouts to fully exploit four‑bit floating‑point (FP4) Tensor Cores. Using this approach, the authors pretrain the Llama‑3‑family 8B model on 160 billion tokens, achieving up to 37.9 K tokens/s/GPU—significantly higher than standard bfloat16 or Transformer Engine FP4 baselines. The study demonstrates that FP4 performance depends on the interplay of scaling, operand packing, layout, and execution path, with downstream task rankings diverging from training‑loss rankings.
arXiv:2609.26333v1 Announce Type: new Abstract: Prefill and decode reward different approaches to quantization: low-precision arithmetic accelerates prompt processing, while compact weights reduce me...
The paper reports a large‑scale post‑training ternarisation of the Qwen3 language model, extending a conversion pipeline from the 4B to the 8B variant. Using KOTMS rotation, E2M‑ATQ adaptive ternarisation, and GPTQ‑style error compensation, the authors achieve a 1.361× perplexity ratio across three corpora and retain 78.5% of the FP16 accuracy on zero‑shot tasks, with the 8B model outperforming the 4B by 8.9 percentage points. The study also demonstrates lossless lattice‑aware packing, producing an 8.24 GiB checkpoint that preserves perplexity, and shows that direct packed execution can reach 15.52 tokens/s in 7.35 GiB, though packed GEMV remains slower than FP16 cuBLAS.
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.
The paper introduces SPECTRA, a runtime‑reconfigurable tiled architecture designed to accelerate speculative decoding for large language models on edge devices. SPECTRA adapts its compute engine within each tile between systolic GEMM execution and vector‑lane GEMV execution, while dynamically adjusting tile count, kernel partitioning, and communication patterns across tiles. Experiments on a 20‑tile FPGA prototype demonstrate up to a 2.09× speedup from tile‑level reconfiguration and an additional 1.25× improvement from system‑level adaptability compared to fixed designs.
The paper presents a Pareto atlas of LLM inference optimizations, mapping cost, quality, and latency trade‑offs for Qwen2.5‑7B‑Instruct on L4, A100, and H100 GPUs. Using 54 measured configurations and a calibrated simulator, it identifies 18 of 36 setups on the Pareto frontier, showing that combined methods outperform single ones. Quality tests reveal that AWQ 4bit and FP8 weights offer significant latency reductions while largely preserving accuracy, but naive FP8 KV caching fails to answer any questions correctly.
arXiv:2610.02598v1 Announce Type: new Abstract: Deploying LLMs on consumer-grade GPUs with insufficient memory to hold their weights can result in prohibitively slow inference, because decoding repea...