arXiv subjects

Md Romyull Islam

Publications and source records attributed to Md Romyull Islam.

5 recordsLinked to original sources

arXiv2026

AQLoRA: A Zero-Search Recipe for Fast Quantized LoRA Fine-Tuning

Quantized fine-tuning (QLoRA) saves memory but not time. It dequantizes every 4-bit weight on the fly, so it trains more slowly than fp16 LoRA. We present AQLoRA (Adaptive-Quantization LoRA), a recipe that buys part of that time back. One CPU pass over the weights sets everything, with no search and no calibration data. The pass ranks layers by NF4 reconstruction error and keeps the top-K in fp16 under a memory budget. Those layers skip dequantization, which is where the speed comes from. A quality setting adapts every layer. A speed setting adapts only the top blocks, so the backward pass stops early. The rule reproduces Unsloth's hand-curated dynamic-4bit selection exactly, in seconds, where search-based allocation needs repeated calibration passes. We evaluate on Commonsense-170K across six models and four architecture families, from 1.4B to 14B. The speed setting trains 11.1 +/- 2.7% faster than well-tuned QLoRA and gives up about one accuracy point. It was faster in all nine independent timing sessions, at worst by 7%. The quality setting trains 4.8 +/- 2.4% faster. Its accuracy is level with QLoRA on every model and within a point of fp16 LoRA, for 0.2 GiB more memory. These error bars are measured between independent sessions, not within one. Earning them taught us three rules for timing on shared hardware. Fix the measurement duration, not the step count. Measure the noise floor from a duplicated arm, not a nearly identical method. Repeat whole sessions: a floor computed inside one sweep understates the real uncertainty several times over, and the random seed controls almost none of it. We validate the recipe with controls and report the two that failed. Choosing adapter layers by weight density is no better than random. Choosing protected layers by quantization error is not either. The count of protected layers, not their identity, carries the speed effect.

cs.LG↗

arXiv2026

NeuroPrefetcher: Storage-Aware Sparse LLM Inference via Delta Prefetching

Deploying large language models on edge devices is increasingly limited by a widening gap between model size and available memory. Existing approaches such as quantization, smaller models, and offloading can raise the effective memory limit, but they still assume that the model can be compressed or partitioned to fit within some budget. We target the harder model-exceeds-memory setting, in which the model remains larger than resident memory throughout execution and storage becomes an active source of weights on the critical path. We observe that MLP activity during autoregressive decoding has strong temporal locality: approximately 82-85% of active neurons persist from one token to the next. This means that most sparse weights needed for the current token are already resident, and only the newly needed rows must be fetched from storage. We present NeuroPrefetcher, a storage-backed LLM inference system that exploits this property through predictive delta prefetching. After layer 0, a single GPU-resident predictor, occupying 2.86% of base model parameters, predicts sparse activity for all downstream MLP layers in one forward pass. The runtime compares these predictions against resident GPU buffers and issues application-scheduled NVMe reads only for incoming delta rows, replacing reactive operating-system demand paging with explicit, model-aware weight movement. On real unified-memory edge hardware, NeuroPrefetcher achieves 7.9-12.0x speedup over llama.cpp across constrained memory budgets.

cs.DC↗

arXiv2025

Characterizing and Understanding Energy Footprint and Efficiency of Small Language Model on Edges

Cloud-based large language models (LLMs) and their variants have significantly influenced real-world applications. Deploying smaller models (i.e., small language models (SLMs)) on edge devices offers additional advantages, such as reduced latency and independence from network connectivity. However, edge devices' limited computing resources and constrained energy budgets challenge efficient deployment. This study evaluates the power efficiency of five representative SLMs - Llama 3.2, Phi-3 Mini, TinyLlama, and Gemma 2 on Raspberry Pi 5, Jetson Nano, and Jetson Orin Nano (CPU and GPU configurations). Results show that Jetson Orin Nano with GPU acceleration achieves the highest energy-to-performance ratio, significantly outperforming CPU-based setups. Llama 3.2 provides the best balance of accuracy and power efficiency, while TinyLlama is well-suited for low-power environments at the cost of reduced accuracy. In contrast, Phi-3 Mini consumes the most energy despite its high accuracy. In addition, GPU acceleration, memory bandwidth, and model architecture are key in optimizing inference energy efficiency. Our empirical analysis offers practical insights for AI, smart systems, and mobile ad-hoc platforms to leverage tradeoffs from accuracy, inference latency, and power efficiency in energy-constrained environments.

cs.DC↗

arXiv2025

A Sparsity Predicting Approach for Large Language Models via Activation Pattern Clustering

Large Language Models (LLMs) exhibit significant activation sparsity, where only a subset of neurons are active for a given input. Although this sparsity presents opportunities to reduce computational cost, efficiently utilizing it requires predicting activation patterns in a scalable manner. However, direct prediction at the neuron level is computationally expensive due to the vast number of neurons in modern LLMs. To enable efficient prediction and utilization of activation sparsity, we propose a clustering-based activation pattern compression framework. Instead of treating each neuron independently, we group similar activation patterns into a small set of representative clusters. Our method achieves up to 79.34% clustering precision, outperforming standard binary clustering approaches while maintaining minimal degradation in perplexity (PPL) scores. With a sufficiently large number of clusters, our approach attains a PPL score as low as 12.49, demonstrating its effectiveness in preserving model quality while reducing computational overhead. By predicting cluster assignments rather than individual neuron states, future models can efficiently infer activation patterns from pre-computed centroids. We detail the clustering algorithm, analyze its effectiveness in capturing meaningful activation structures, and demonstrate its potential to improve sparse computation efficiency. This clustering-based formulation serves as a foundation for future work on activation pattern prediction, paving the way for efficient inference in large-scale language models.

cs.LG↗

arXiv2024

Activation Sparsity Opportunities for Compressing General Large Language Models

Deploying local AI models, such as Large Language Models (LLMs), to edge devices can substantially enhance devices' independent capabilities, alleviate the server's burden, and lower the response time. Owing to these tremendous potentials, many big tech companies have released several lightweight Small Language Models (SLMs) to bridge this gap. However, we still have huge motivations to deploy more powerful (LLMs) AI models on edge devices and enhance their smartness level. Unlike the conventional approaches for AI model compression, we investigate activation sparsity. The activation sparsity method is orthogonal and combinable with existing techniques to maximize the compression rate while maintaining great accuracy. LLMs' Feed-Forward Network (FFN) components, which typically comprise a large proportion of parameters (around 2/3), ensure that our FFN optimizations would have a better chance of achieving effective compression. Moreover, our findings are beneficial to general LLMs and are not restricted to ReLU-based models. This work systematically investigates the tradeoff between enforcing activation sparsity and perplexity (accuracy) on state-of-the-art LLMs. Our empirical analysis demonstrates that we can obtain around 50% of main memory and computing reductions for critical FFN components with negligible accuracy degradation. This extra 50% sparsity does not naturally exist in the current LLMs, which require tuning LLMs' activation outputs by injecting zero-enforcing thresholds. To obtain the benefits of activation sparsity, we provide a guideline for the system architect for LLM prediction and prefetching. The success prediction allows the system to prefetch the necessary weights while omitting the inactive ones and their successors, therefore lowering cache and memory pollution and reducing LLM execution time on resource-constrained edge devices.

cs.LG↗