arXiv subjects

Prabod Rathnayaka

Publications and source records attributed to Prabod Rathnayaka.

5 recordsLinked to original sources

arXiv2026

BaseRT: Advancing Best-in-Class LLM Inference with Apple M5 Neural Accelerators

Apple's M5 generation introduces a redesigned GPU architecture in which every core carries a dedicated Neural Accelerator: on-die matrix units exposed through the Metal~4 tensor API. We show that BaseRT, our native Metal inference runtime for large language models on Apple Silicon, exploits these units to push inference throughput on Apple hardware substantially beyond both llama.cpp and MLX. Building on BaseRT's framework-free design, we add a family of hand-written Metal~4 tensor-core kernels (including dense and mixture-of-experts GEMM and flash-attention prefill kernels) that route the compute-bound matrix multiplications of inference through the M5 Neural Accelerators while leaving the memory-bound decode path on our existing specialised kernels. On an Apple M5 Pro, across fifteen model configurations spanning the Qwen3, Qwen3.5/3.6, Llama~3.2, and Gemma~4 families from sub-1B to 35B parameters, BaseRT delivers up to $6.4\times$ higher prompt-processing throughput than llama.cpp and $3.9\times$ higher than MLX, with the largest margins on the mixture-of-experts models where matrix multiplication dominates, while maintaining its lead on decode of up to $1.75\times$ over llama.cpp and $1.33\times$ over MLX. These results establish a new performance ceiling for on-device LLM inference and show that the M5's tensor cores are the decisive lever for prompt processing on Apple Silicon. BaseRT is publicly available at https://github.com/basecompute/baseRT.

cs.AR↗

arXiv2026

BaseRT: Best-in-Class LLM Inference on Apple Silicon via Native Metal

We present BaseRT, a native Metal inference runtime for large language models (LLMs) on Apple Silicon, and report the highest inference throughput on this hardware to date. Existing runtimes, including llama.cpp and MLX-based frameworks, incur overhead from abstractions not designed for Metal's execution model or Apple Silicon's unified memory topology. By building natively on Metal with chip-specific kernel fusion, unified memory-aware optimisation, and custom dispatch logic, BaseRT recovers performance that framework-based approaches leave on the table. BaseRT supports a wide range of model families across eight quantisation formats (Q2 to FP16) on all Apple M-series devices. In this paper, we evaluate the Qwen3, Llama 3.2, and Gemma 4 families at Q4 and Q8 quantisation on M3 and M4 Pro devices. BaseRT achieves up to 1.56x higher decode throughput than llama.cpp and up to 1.35x higher than MLX, with substantially larger margins on prefill for mixture-of-experts models, delivering consistent best-in-class throughput from sub-1B to 30B parameter models. These results establish Apple Silicon as a more capable inference platform than previously reported, with direct implications for the emerging edge inference paradigm: as privacy requirements, latency constraints, and cloud cost pressures drive inference toward on-device deployment, performance-optimised local runtimes are a critical enabling layer for this transition. BaseRT is publicly available at https://github.com/basecompute/baseRT

cs.CL↗

arXiv2019

Gated Recurrent Neural Network Approach for Multilabel Emotion Detection in Microblogs

People express their opinions and emotions freely in social media posts and online reviews that contain valuable feedback for multiple stakeholders such as businesses and political campaigns. Manually extracting opinions and emotions from large volumes of such posts is an impossible task. Therefore, automated processing of these posts to extract opinions and emotions is an important research problem. However, human emotion detection is a challenging task due to the complexity and nuanced nature. To overcome these barriers, researchers have extensively used techniques such as deep learning, distant supervision, and transfer learning. In this paper, we propose a novel Pyramid Attention Network (PAN) based model for emotion detection in microblogs. The main advantage of our approach is that PAN has the capability to evaluate sentences in different perspectives to capture multiple emotions existing in a single text. The proposed model was evaluated on a recently released dataset and the results achieved the state-of-the-art accuracy of 58.9%.

cs.CL↗

arXiv2018

Enhancing Decision Making Capacity in Tourism Domain Using Social Media Analytics

Social media has gained an immense popularity over the last decade. People tend to express opinions about their daily encounters on social media freely. These daily encounters include the places they traveled, hotels or restaurants they have tried and aspects related to tourism in general. Since people usually express their true experiences on social media, the expressed opinions contain valuable information that can be used to generate business value and aid in decision-making processes. Due to the large volume of data, it is not a feasible task to manually go through each and every item and extract the information. Hence, we propose a social media analytics platform which has the capability to identify discussion pathways and aspects with their corresponding sentiment and deeper emotions using machine learning techniques and a visualization tool which shows the extracted insights in a comprehensible and concise manner. Identified topic pathways and aspects will give a decision maker some insight into what are the most discussed topics about the entity whereas associated sentiments and emotions will help to identify the feedback.

cs.IR↗

arXiv2018

Sentylic at IEST 2018: Gated Recurrent Neural Network and Capsule Network Based Approach for Implicit Emotion Detection

In this paper, we present the system we have used for the Implicit WASSA 2018 Implicit Emotion Shared Task. The task is to predict the emotion of a tweet of which the explicit mentions of emotion terms have been removed. The idea is to come up with a model which has the ability to implicitly identify the emotion expressed given the context words. We have used a Gated Recurrent Neural Network (GRU) and a Capsule Network based model for the task. Pre-trained word embeddings have been utilized to incorporate contextual knowledge about words into the model. GRU layer learns latent representations using the input word embeddings. Subsequent Capsule Network layer learns high-level features from that hidden representation. The proposed model managed to achieve a macro-F1 score of 0.692.

cs.CL↗