Sökning: "Quantization Aware Training"

Hittade 3 uppsatser innehållade orden Quantization Aware Training.

1. 1. Utilizing energy-saving techniques to reduce energy and memory consumption when training machine learning models : Sustainable Machine Learning

   Master-uppsats, KTH/Skolan för elektroteknik och datavetenskap (EECS)

   Författare :Khalid El Yaacoub; [2024]
   Nyckelord :Sustainable AI; Machine learning; Quantization-Aware Training; Model Distillation; Quantized Distillation; Siamese Neural Networks; Continual Learning; Experience Replay; Data Efficient AI; Energy Consumption; Energy-Savings; Sustainable ML; Computation resources; Hållbar maskin inlärning; Hållbar AI; Maskininlärning; Quantization-Aware Training; Model Distillation; Quantized Distillation; siamesiska neurala nätverk; Continual Learning; Experience Replay; Dataeffektiv AI; Energiförbrukning; Energibesparingar; Beräkningsresurser;

   Sammanfattning : Emerging machine learning (ML) techniques are showing great potential in prediction performance. However, research and development is often conducted in an environment with extensive computational resources and blinded by prediction performance. LÄS MER

3. 2. Mixed Precision Quantization for Computer Vision Tasks in Autonomous Driving

   Master-uppsats, KTH/Skolan för elektroteknik och datavetenskap (EECS)

   Författare :Sri Janani Rengarajan; [2022]
   Nyckelord :Quantization; Neural Networks; Quantization Aware Training; Mixed precision; Semantic segmentation; Hessian; Kvantisering; Neurala nätverk; Kvantiseringsmedveten träning; Blandad precision; Semantisk segmentering; Hessian;

   Sammanfattning : Quantization of Neural Networks is popular technique for adopting computation intensive Deep Learning applications to edge devices. In this work, low bit mixed precision quantization of FPN-Resnet18 model trained for the task of semantic segmentation is explored using Cityscapes and Arriver datasets. LÄS MER

4. 3. QPLaBSE: Quantized and Pruned Language-Agnostic BERT Sentence Embedding Model : Production-ready compression for multilingual transformers

   Master-uppsats, KTH/Skolan för elektroteknik och datavetenskap (EECS)

   Författare :Sarthak Langde; [2021]
   Nyckelord :Transformers; LaBSE; Quantization; Pruning; PyTorch; TensorFlow; ONNX; Transformatorer; LaBSE; Kvantisering; Beskärning; PyTorch; TensorFlow; ONNX;

   Sammanfattning : Transformer models perform well on Natural Language Processing and Natural Language Understanding tasks. Training and fine-tuning of these models consume a large amount of data and computing resources. Fast inference also requires high-end hardware for user-facing products. LÄS MER