arXiv subjects

Deval Shah

Publications and source records attributed to Deval Shah.

6 recordsLinked to original sources

arXiv2025

FarSkip-Collective: Unhobbling Blocking Communication in Mixture of Experts Models

Blocking communication presents a major hurdle in running MoEs efficiently in distributed settings. To address this, we present FarSkip-Collective which modifies the architecture of modern models to enable overlapping of their computation with communication. Our approach modifies the architecture to skip connections in the model and it is unclear a priori whether the modified model architecture can remain as capable, especially for large state-of-the-art models and while modifying all of the model layers. We answer this question in the affirmative and fully convert a series of state-of-the-art models varying from 16B to 109B parameters to enable overlapping of their communication while achieving accuracy that is comparable with their original open-source releases. For example, we convert Llama 4 Scout (109B) via self-distillation and achieve average accuracy within 1% of its instruction tuned release averaged over a wide range of downstream evaluations. In addition to demonstrating retained accuracy of the large modified models, we realize the benefits of FarSkip-Collective through optimized implementations that explicitly overlap communication with computation, accelerating both training and inference in existing frameworks. For inference, we demonstrate 32.6% speedup in Time To First Token when serving a converted DeepSeek-V3 architecture with expert parallelism in SGLang and achieve 97.3% communication-computation overlap during the prefill stage. During training, our approach enables 88.9% communication overlap of the all-to-all communication collectives when pre-training DeepSeek-V3 MoE layers with expert parallelism.

cs.LG↗

arXiv2025

BlindSight: Harnessing Sparsity for Efficient Vision-Language Models

Large vision-language models (VLMs) enable joint processing of text and images. However, incorporating vision data significantly increases the prompt length, resulting in a longer time to first token (TTFT). This bottleneck can be alleviated by leveraging the inherent sparsity in the attention computation. Analyzing these attention patterns in VLMs when processing a series of images, we observe the absence of inter-image attention in a substantial portion of layers. Based on this, we propose BlindSight: an approach to optimize multi-image VLM inference using an input-template-aware attention sparsity mask with no runtime overhead. We utilize a dataset to derive a prompt-agnostic categorization for attention heads: Dense, Sink, Intra-Image, and Intra-Image+Sink. We develop a Triton-based GPU kernel to leverage this sparsity. BlindSight achieves a 1.8-3.2x speedup in the attention computation (prompt length 36K-300K). BlindSight generalizes across VLMs (Qwen2-VL, Qwen2.5-VL, Gemma 3), with only a 0.78% absolute accuracy degradation on average on multi-image comprehension benchmarks. Finally, we advocate for the design of efficient VLMs that combine BlindSight-inspired sparse and dense layers.

cs.CV↗

arXiv2023

Learning Label Encodings for Deep Regression

Deep regression networks are widely used to tackle the problem of predicting a continuous value for a given input. Task-specialized approaches for training regression networks have shown significant improvement over generic approaches, such as direct regression. More recently, a generic approach based on regression by binary classification using binary-encoded labels has shown significant improvement over direct regression. The space of label encodings for regression is large. Lacking heretofore have been automated approaches to find a good label encoding for a given application. This paper introduces Regularized Label Encoding Learning (RLEL) for end-to-end training of an entire network and its label encoding. RLEL provides a generic approach for tackling regression. Underlying RLEL is our observation that the search space of label encodings can be constrained and efficiently explored by using a continuous search space of real-valued label encodings combined with a regularization function designed to encourage encodings with certain properties. These properties balance the probability of classification error in individual bits against error correction capability. Label encodings found by RLEL result in lower or comparable errors to manually designed label encodings. Applying RLEL results in 10.9% and 12.4% improvement in Mean Absolute Error (MAE) over direct regression and multiclass classification, respectively. Our evaluation demonstrates that RLEL can be combined with off-the-shelf feature extractors and is suitable across different architectures, datasets, and tasks. Code is available at https://github.com/ubc-aamodt-group/RLEL_regression.

cs.LG↗

arXiv2022

Label Encoding for Regression Networks

Deep neural networks are used for a wide range of regression problems. However, there exists a significant gap in accuracy between specialized approaches and generic direct regression in which a network is trained by minimizing the squared or absolute error of output labels. Prior work has shown that solving a regression problem with a set of binary classifiers can improve accuracy by utilizing well-studied binary classification algorithms. We introduce binary-encoded labels (BEL), which generalizes the application of binary classification to regression by providing a framework for considering arbitrary multi-bit values when encoding target values. We identify desirable properties of suitable encoding and decoding functions used for the conversion between real-valued and binary-encoded labels based on theoretical and empirical study. These properties highlight a tradeoff between classification error probability and error-correction capabilities of label encodings. BEL can be combined with off-the-shelf task-specific feature extractors and trained end-to-end. We propose a series of sample encoding, decoding, and training loss functions for BEL and demonstrate they result in lower error than direct regression and specialized approaches while being suitable for a diverse set of regression problems, network architectures, and evaluation metrics. BEL achieves state-of-the-art accuracies for several regression benchmarks. Code is available at https://github.com/ubc-aamodt-group/BEL_regression.

cs.LG↗

arXiv2021

Characterizing and Improving the Resilience of Accelerators in Autonomous Robots

Motion planning is a computationally intensive and well-studied problem in autonomous robots. However, motion planning hardware accelerators (MPA) must be soft-error resilient for deployment in safety-critical applications, and blanket application of traditional mitigation techniques is ill-suited due to cost, power, and performance overheads. We propose Collision Exposure Factor (CEF), a novel metric to assess the failure vulnerability of circuits processing spatial relationships, including motion planning. CEF is based on the insight that the safety violation probability increases with the surface area of the physical space exposed by a bit-flip. We evaluate CEF on four MPAs. We demonstrate empirically that CEF is correlated with safety violation probability, and that CEF-aware selective error mitigation provides 12.3x, 9.6x, and 4.2x lower Failures-In-Time (FIT) rate on average for the same amount of protected memory compared to uniform, bit-position, and access-frequency-aware selection of critical data. Furthermore, we show how to employ CEF to enable fault characterization using 23,000x fewer fault injection (FI) experiments than exhaustive FI, and evaluate our FI approach on different robots and MPAs. We demonstrate that CEF-aware FI can provide insights on vulnerable bits in an MPA while taking the same amount of time as uniform statistical FI. Finally, we use the CEF to formulate guidelines for designing soft-error resilient MPAs.

cs.AR↗

arXiv2018

Analyzing Machine Learning Workloads Using a Detailed GPU Simulator

Most deep neural networks deployed today are trained using GPUs via high-level frameworks such as TensorFlow and PyTorch. This paper describes changes we made to the GPGPU-Sim simulator to enable it to run PyTorch by running PTX kernels included in NVIDIA's cuDNN library. We use the resulting modified simulator, which has been made available publicly with this paper, to study some simple deep learning workloads. With our changes to GPGPU-Sim's functional simulation model, we find GPGPU-Sim performance model running a cuDNN enabled implementation of LeNet for MNIST reports results within 30% of real hardware. Using GPGPU-Sim's AerialVision performance analysis tool we observe that cuDNN API calls contain many varying phases and appear to include potentially inefficient microarchitecture behaviour such as DRAM partition bank camping, at least when executed on GPGPU-Sim's current performance model.

cs.DC↗