Searcharxiv⌕ Search

arXiv subjects

Saptarshi Mitra

Publications and source records attributed to Saptarshi Mitra.

9 recordsLinked to original sources

MOSAIC: Efficient Mixture-of-Agent Scheduling via Adaptive Aggregation and Inference Concurrency

Mixture-of-Agents (MoA) systems improve reasoning accuracy by routing each query to multiple expert LLMs and aggregating their outputs. Efficiently executing this workload on limited GPU resources has bottlenecks. Skill-based routing creates skewed expert demand, and combining instruction-tuned LLMs with long-reasoning models results in extreme variability in generation lengths. Consequently, traditional scheduling strategies suffer from significant GPU idling and throughput collapse due to load imbalances. We present MOSAIC, a scheduling framework for accelerating batch-oriented MoA workloads. MOSAIC uses two-stage workload-aware scheduling: an expert placement stage, where an integer linear program (ILP) decides expert placement, followed by a query-sequencing stage that schedules requests based on predicted output lengths. Additionally, MOSAIC uses confidence-aware adaptive aggregation, leveraging inter-expert agreement to bypass the heavy aggregator LLM for consensus queries. On a 4-GPU system, MOSAIC achieves 2.7-3.4x end-to-end speedups over round-robin scheduling and reduces the mean time-to-answer by 3.9-7.8x, while matching accuracy within +/-1pp.

cs.LG↗

LowRank-SSM: Hardware-Software Co-Design for Rank-Reduced Mamba Acceleration on FPGA

State Space Models (SSMs) such as Mamba and Mamba-2 achieve linear-time autoregressive inference, making them attractive for latency-sensitive and resource-constrained deployment. Yet every generated token still reads the full input and output projection weights of every layer from off-chip memory, and on an FPGA these two projections take the majority of a layer's cycles. Existing accelerators lower this cost by quantizing the weights or by speeding up the scan, but none reduce the number of weight elements a token must read, and a low-rank factorization on its own does not help: without a datapath built for it, the two-stage product takes as long as the dense one. We present \textbf{LowRank-SSM}, a hardware-software co-design framework that jointly addresses three constraints of rank-reduced Mamba inference: where rank should be spent, how the resulting factors should execute, and how reduced weight traffic should be delivered by the FPGA memory system. A sensitivity-guided per-projection search allocates rank under a perplexity ceiling while restricting compression to Mamba-2 operands that tolerate approximation. A rank-configurable projection datapath maps both factor stages onto the existing GEMV engines, keeping the intermediate vector on chip and changing rank through loop bounds and weight images rather than bitstream recompilation. A platform-aware NoC/AXI memory subsystem then balances and prefetches weight streams to sustain 42\,GB/s. Together, these software and hardware choices convert structural rank reduction into measured latency reduction: in our ablation, 93\% of the projection traffic removed by rank reduction appears as saved token latency. With a rank schedule found under a 5\% perplexity ceiling, which costs 1.6 points of average zero-shot accuracy, the complete accelerator decodes Mamba2-2.7B at 16.79 tokens/s and 0.41\,J per token.

cs.AR↗

RaReCache: Bridging the Gap in Cross-Model KV Cache Reuse via Rank disagreement-based Selective Recomputation

Cross-model KV-cache reuse remains a key challenge in modern LLM serving. Coding agents and multi-model systems increasingly route a shared context across models: a user may switch models mid-session, or a cascade may escalate a difficult query. Because KV caches contain model-specific representations, each switch typically forces the receiving model to prefill the entire context from scratch. Recent work shows that closed-form linear maps can translate KV caches between models in the same family, but transfer accuracy degrades as the model-size gap widens. In this paper, we establish that these transfer failures are concentrated in a small subset of information-dense tokens. To bridge this gap, we introduce RaReCache, a framework that enables a large target model to decode accurately from a cache prefilled by a much smaller source via selective recomputation. RaReCache identifies these critical positions using a novel rank disagreement metric, scoring each token by the energy of its mapped KV in output directions weakly supported by the calibration data. Across two model families and five benchmarks, on a 23x parameter gap (Qwen3-0.6B to 14B) recomputing just 30% of positions retains 95-99% of the target accuracy, whereas on a 8.8x gap (Llama3-8B to 70B), recomputing 40% retains 96.5% of the target accuracy. RaReCache largely removes sensitivity to source-model size, and achieves up to a 3.04x prefill speedup. For online serving, it handles 1.8x the request throughput of target prefill on a single GPU, and at the target's saturation load, reduces median and 99th-percentile time-to-first-token (TTFT) by 5.0x and 6.4x respectively, with a 30% recompute budget. RaReCache establishes an efficient serving paradigm where small models prefill on behalf of massive targets, enabling large models to recompute only critical tokens, drastically reducing prefill latency.

cs.AI↗

ETA: Exit on Tool Arrival in Speculative Agent Loops

Tool-using language agents in explicit thinking modes generate long chains of thought before acting, making each turn inherently serial: later tokens depend on the tool observation. Recent speculative tool dispatch methods overlap tool execution with reasoning, but offers limited benefit when reasoning dominates latency. We present ETA, that treats the speculative fork not as an actuator but as a sensor of the model's decision state, using it to terminate the reasoning of the main decode. It truncates reasoning only when the prediction is stable, agrees with the dispatched call, passes a free probe measuring whether a call will be emitted at all, and the tool result has arrived: a deadline that bounds per-turn regret by the remaining decode cost. ETA achieves speedups of 1.33x on tau2 airline/retail, 1.26x on tau2 telecom, and 1.13x on BFCL v4 multi-turn. On tau2, it decodes 24% fewer tokens at a nearly unchanged per-token rate without dropping action quality.

cs.SE↗

RAPID-LLM: Resilience-Aware Performance analysis of Infrastructure for Distributed LLM Training and Inference

RAPID-LLM is a unified performance modeling framework for distributed large language model (LLM) training and inference on GPU clusters, without relying on deployment-specific traces or expensive cycle-level simulation for exploration. From a workload and hardware specification, it builds hardware-aware operator-level execution models that capture tiling, memory-hierarchy effects, communication, and memory feasibility under hybrid parallelism. Its backend simulates explicit multidimensional interconnects with congestion-aware routing and support for degraded and failed links, enabling scalable what-if analysis across topology, mapping, and hardware design choices. Across 124 evaluation cases spanning inference and dense, fully sharded, and mixture-of-experts training on A100 and H100 GPUs, RAPID-LLM achieves an overall mean absolute percentage error (MAPE) of 10.0\%. Its network predictions stay within 8\% of ns-3 on representative communication patterns. Case studies demonstrate how RAPID-LLM enables fast, systematic sweeps over hybrid-parallel configurations, quantifies sensitivity to link faults under realistic routing and congestion, and evaluates hypothetical GPU design variants including 3D-stacked HBM-on-GPU scenarios.

cs.PF↗

Characterizing State Space Model and Hybrid Language Model Performance with Long Context

Emerging applications such as AR are driving demands for machine intelligence capable of processing continuous and/or long-context inputs on local devices. However, currently dominant models based on Transformer architecture suffers from the quadratic computational and memory overhead, which hinders applications required to process long contexts. This has spurred a paradigm shift towards new architectures like State Space Models (SSMs) and SSM-Transformer hybrid models, which provide near-linear scaling. The near-linear scaling enabled efficient handling of millions of tokens while delivering high performance in recent studies. Although such works present promising results, their workload characteristics in terms of computational performance and hardware resource requirements are not yet thoroughly explored, which limits our understanding of their implications to the system level optimizations. To address this gap, we present a comprehensive, compara-ive benchmarking of carefully selected Transformers, SSMs, and hybrid models specifically for long-context inference on consumer and embedded GPUs. Our analysis shows that SSMs are well-suited for on-device AI on consumer and embedded GPUs for long context inferences. While Transformers are up to 1.9x faster at short sequences (<8K tokens), SSMs demonstrate a dramatic performance inversion, becoming up to 4x faster at very long contexts (~57K tokens), thanks to their linear computational complexity and ~64% reduced memory footrprint. Our operator-level analysis reveals that custom SSM kernels like selective scan despite being hardware-aware to minimize memory IO, dominate the inference runtime on edge platforms, accounting for over 55% of latency due to their sequential, element-wise nature. SSM-Scope is open-sourced at https://github.com/sapmitra/ssm-scope

cs.AR↗

DeepliteRT: Computer Vision at the Edge

The proliferation of edge devices has unlocked unprecedented opportunities for deep learning model deployment in computer vision applications. However, these complex models require considerable power, memory and compute resources that are typically not available on edge platforms. Ultra low-bit quantization presents an attractive solution to this problem by scaling down the model weights and activations from 32-bit to less than 8-bit. We implement highly optimized ultra low-bit convolution operators for ARM-based targets that outperform existing methods by up to 4.34x. Our operator is implemented within Deeplite Runtime (DeepliteRT), an end-to-end solution for the compilation, tuning, and inference of ultra low-bit models on ARM devices. Compiler passes in DeepliteRT automatically convert a fake-quantized model in full precision to a compact ultra low-bit representation, easing the process of quantized model deployment on commodity hardware. We analyze the performance of DeepliteRT on classification and detection models against optimized 32-bit floating-point, 8-bit integer, and 2-bit baselines, achieving significant speedups of up to 2.20x, 2.33x and 2.17x, respectively.

cs.LG↗

YOLOBench: Benchmarking Efficient Object Detectors on Embedded Systems

We present YOLOBench, a benchmark comprised of 550+ YOLO-based object detection models on 4 different datasets and 4 different embedded hardware platforms (x86 CPU, ARM CPU, Nvidia GPU, NPU). We collect accuracy and latency numbers for a variety of YOLO-based one-stage detectors at different model scales by performing a fair, controlled comparison of these detectors with a fixed training environment (code and training hyperparameters). Pareto-optimality analysis of the collected data reveals that, if modern detection heads and training techniques are incorporated into the learning process, multiple architectures of the YOLO series achieve a good accuracy-latency trade-off, including older models like YOLOv3 and YOLOv4. We also evaluate training-free accuracy estimators used in neural architecture search on YOLOBench and demonstrate that, while most state-of-the-art zero-cost accuracy estimators are outperformed by a simple baseline like MAC count, some of them can be effectively used to predict Pareto-optimal detection models. We showcase that by using a zero-cost proxy to identify a YOLO architecture competitive against a state-of-the-art YOLOv8 model on a Raspberry Pi 4 CPU. The code and data are available at https://github.com/Deeplite/deeplite-torch-zoo

cs.CV↗

DeepGEMM: Accelerated Ultra Low-Precision Inference on CPU Architectures using Lookup Tables

A lot of recent progress has been made in ultra low-bit quantization, promising significant improvements in latency, memory footprint and energy consumption on edge devices. Quantization methods such as Learned Step Size Quantization can achieve model accuracy that is comparable to full-precision floating-point baselines even with sub-byte quantization. However, it is extremely challenging to deploy these ultra low-bit quantized models on mainstream CPU devices because commodity SIMD (Single Instruction, Multiple Data) hardware typically supports no less than 8-bit precision. To overcome this limitation, we propose DeepGEMM, a lookup table based approach for the execution of ultra low-precision convolutional neural networks on SIMD hardware. The proposed method precomputes all possible products of weights and activations, stores them in a lookup table, and efficiently accesses them at inference time to avoid costly multiply-accumulate operations. Our 2-bit implementation outperforms corresponding 8-bit integer kernels in the QNNPACK framework by up to 1.74x on x86 platforms.

cs.LG↗