Author丨Mialo@Zhihu Source丨https://zhuanlan.zhihu.com/p/486360176 1. Background Introduction Analyzing the PyTorch memory management mechanism primarily aims to reduce the impact of “memory fragmentation”. A simple example is as follows: As shown in the figure above, suppose we want to allocate 800MB of memory. Although the total free memory is 1000MB, the free memory shown in the upper figure … Read more
“Standing on the shoulders of giants” — this is precisely the benefit that pre-trained models bring in deep learning. Training deep neural networks from scratch is both resource-intensive and time-consuming. However, the good news is that PyTorch’s pre-trained models can effectively solve this issue. PyTorch provides models that have learned robust features on large datasets, … Read more
1. Two Parallelization Models in Torch 1.1 DataParallel DataParallel is a data parallel method provided by PyTorch for model training on multiple GPUs within a single machine. It achieves parallel computation by splitting the input data into several sub-parts (mini-batches) and distributing these sub-parts to different GPUs. During the forward pass, the input data is … Read more
Introduction When constructing models, especially when dealing with the input-output data formats between layers, some commonly used data processing functions such as tensor calculations and broadcasting mechanisms are very important. They remain indispensable when later using the Transformers library with pre-trained models. This article aims to explain the most commonly used Pytorch processing functions for … Read more
Click on the above “Beginner’s Visual Learning” to select “Star” or “Pin“ Important content delivered promptly TorchGAN is a GAN design and development framework based on PyTorch. This framework aims to provide construction modules for popular GANs and allows customization for cutting-edge research. Using the modular structure of TorchGAN, you can: Experiment with popular GAN … Read more
Case Introduction Generative Adversarial Networks (GANs) are a type of deep learning model consisting of a generator network and a discriminator network. They improve their capabilities through adversarial training, competing against each other. The generator network attempts to produce samples that resemble real data, while the discriminator network tries to distinguish between samples generated by … Read more
1. Introduction to GAN “Foodie, food spirit, foodies are the best of people”. This GAN foodie is not the same as that foodie. The GAN we are going to discuss is the Generative Adversarial Network proposed by Goodfellow in 2014. So what is so magical about GAN? Conventional deep learning tasks such as image classification, … Read more
Follow the public account “ML_NLP” Set as “Starred”, heavy content delivered first! Author & Code: Harbin Institute of Technology SCIR Shi Jihao 1. Introduction 1.1 Article Organization This article briefly introduces the basic principles of CNN and uses the sentence-level sentiment classification task as an example to introduce the use of CNN methods for feature … Read more
Source: Deephub Imba This article contains approximately 3000 words, and is suggested to be read in 5 minutes This article will introduce the application of CNN to solve simple 2D path planning problems. Convolutional Neural Networks (CNN) are popular models for tasks such as image classification, segmentation, and object detection. This article applies CNN to … Read more
About 3400 words, recommended reading time: 10+ minutes This article will comprehensively introduce the attention mechanism in CNNs, from basic concepts to practical implementations, providing readers with in-depth understanding and practical guidance. The attention mechanism has become an indispensable component of deep learning models, especially in Convolutional Neural Networks (CNNs). By enabling models to selectively … Read more