The paper presents a method for compiling multi‑threshold operators that support quantized, non‑monotonic, and fine‑granular activation functions in custom DNN accelerators. It extends the multi‑threshold representation to multi‑dimensional thresholds, integrates algebraic conversion rules into an ONNX‑based compiler, and validates the approach on both monotonic and non‑monotonic functions using the FINN framework. The authors also characterize hardware costs compared to graph‑level decomposition and discuss potential use in quantization‑aware training.
arXiv:2602. 22352v2 Announce Type: replace-cross Abstract: With the continuous growth of neural network scales, low-precision quantization is widely used in edge accelerators.
arXiv:2607. 03148v1 Announce Type: cross Abstract: Activation functions are considered an essential primitive for neural nonlinearity, i.
arXiv:2608. 06916v1 Announce Type: new Abstract: Quantized Neural Networks~(QNN) with low-bitwidth data have proven promising in efficient storage and computation on edge devices.
arXiv:2602. 23334v2 Announce Type: replace-cross Abstract: Neural network accelerators have been widely applied to edge devices for complex tasks like object tracking, image recognition, etc.
The paper introduces the Quantization Analysis Tool, a system built on the ONNX framework that streamlines quantization workflows for deep learning models. It offers layer‑wise sensitivity analysis, visualizations of weight and activation distributions, and guidance for selecting precision levels to balance model size, latency, and accuracy. Experiments on various neural network architectures show that the tool improves quantized accuracy and overall deployment efficiency.
DualQuant introduces a dual‑scale, calibration‑free post‑training quantization framework that jointly optimizes weight‑reconstruction MSE and derives per‑channel activation rescalings, enabling effective 4‑bit weight and activation quantization without calibration data. The method employs a non‑differentiable optimization solved via variable splitting and block coordinate descent, supporting a wide range of formats including MX and NVFP4. Experiments on Llama and Qwen models show DualQuant matches or surpasses calibration‑based SmoothQuant and significantly outperforms other calibration‑free baselines, achieving up to 2.61× memory reduction and 2.46× throughput speedup on an RTX 4090.
The paper introduces RAMP, a method for robust adaptive mixed‑precision quantization of vision models on edge CPUs. It evaluates 13 sensitivity metrics across four neural networks, finding that Jensen‑Shannon Divergence consistently identifies layers that can be safely quantized. Using K‑Means clustering on these metrics, RAMP achieves near‑lossless accuracy with an average 1.81× speed‑up, while cautioning against excluding low‑speed‑up layers that can fragment the computational graph.
Achieving nanosecond-scale inference latency for deep neural networks (DNNs) has become a primary architectural concern for latency-critical applications. While Field-Programmable Gate Arrays (FPGAs) offer a promising substrate for low-latency inference, conventional FPGA accelerators remain arithmetic-centric, using LUTs primarily as building blocks for numerical operators and peripheral logic.
In this paper, we present CAT-Q, Cost-efficient and Accurate Ternary Quantization, for compressing and accelerating LLMs. Unlike existing state-of-the-art ternary quantization methods that rely on data-intensive and costly quantization-aware training to mitigate severe performance degradation, CAT-Q is a simple yet effective post-training quantization scheme that is readily applicable to LLMs with diverse architectures and model sizes.
FAME is an FPGA-based platform that evaluates approximate multipliers directly in hardware, eliminating slow CPU/GPU LUT emulation and reducing evaluation time for DNN inference. It also introduces a pattern-guided retraining method that uses multiplier-specific patterns to recover accuracy losses. Experiments on ResNet‑18 and MobileNetV2 over ImageNet show up to 3.47× faster multiplier evaluation and a 65.5% accuracy improvement over prior retraining approaches.
arXiv:2606. 26650v1 Announce Type: cross Abstract: In this paper, we present CAT-Q, Cost-efficient and Accurate Ternary Quantization, for compressing and accelerating LLMs.
Deploying deep learning models on edge CPUs is bottlenecked by computational and memory constraints. Mixed-precision quantization promises to reduce inference latency while preserving accuracy. Howeve...