This article presents practical strategies for building high‑performance PyTorch training pipelines, covering bottleneck identification, efficient data loading, RAM‑based datasets, profiling tools, multi‑GPU training with DataParallel and DistributedDataParallel, custom loss implementation, and hardware‑vs‑software trade‑offs to accelerate deep‑learning workloads.
This article details the background, MMoE model architecture, baseline setup, environment configuration, data preprocessing, training process, evaluation results, and department information for the 58.com recruitment recommendation AI competition using the WPAI platform.
This article explains how PyTorch's BP Engine dynamically builds the computation graph for back‑propagation, detailing its C++ class structure, thread management, task queues, and key functions such as start_threads, compute_dependencies, execute, and evaluate_function, with illustrative code examples.
This article explains how PyTorch implements its dynamic autograd graph using the core C++ classes Variable, Function, and Engine, detailing their initialization, inheritance hierarchy, code structures, and the execution flow of backward propagation.
Bagua is a distributed training framework co‑developed by Kuaishou and ETH Zürich that combines algorithmic and system‑level optimizations—such as decentralized, asynchronous, and compressed communication—to achieve up to 60% higher performance than existing frameworks like PyTorch‑DDP, Horovod, and BytePS across various AI workloads.
An in‑depth comparison of TensorFlow 2.x and PyTorch 1.8 highlights new features, deployment options like TensorFlow Lite and PyTorch Mobile, coding style differences, and practical guidance on choosing the right deep‑learning library for various projects and skill levels.
This article explains the motivation behind residual networks, describes the architecture of ResNet including residual blocks and skip connections, lists available Keras implementations, and provides a step‑by‑step PyTorch tutorial with complete code to construct and test ResNet‑50/101/152 models.
This article reviews the latest releases of PyTorch 1.8 and TensorFlow 2.5, outlining their new functionalities, ecosystem tools such as TensorFlow.js, Lite, and TFX, as well as PyTorch Mobile and Lightning, and provides guidance on selecting the most suitable framework for different deep‑learning projects.
This tutorial explains how to install, launch, and use TensorBoard and Visdom with PyTorch, providing step‑by‑step commands, code examples for logging training metrics, and visualizing images and plots to monitor deep‑learning experiments.
This article explains how Tubi leverages machine learning—particularly PyTorch, Databricks, and cloud services—to improve content understanding, advertising technology, and recommendation systems within its advertising‑based video‑on‑demand platform, outlining the three AVOD pillars, technical stack, and future research directions.
This article explains how to build a code clone detection service using a CodeBERT classification model, create a custom TorchServe handler, package the model with torch-model-archiver, launch the service, and test it with example code pairs to demonstrate clone and non‑clone predictions.
This developer community weekly roundup highlights CCTV's new big‑data governance platform, RedMonk's programming language rankings, Chromium‑based browsers adopting a four‑week release cycle, PyTorch 1.8 with AMD support, the world’s first AI‑driven earthquake monitoring system, Red Hat OpenShift 4.7, a deep meta‑learning model for city sales prediction, and a CVPR breakthrough in controllable human image generation.
PyTorch 1.8, released by the PyTorch team, bundles over 3,000 commits since 1.7, introducing AMD ROCm support, enhanced Python function conversion, stable FFT and linear‑algebra APIs, complex‑tensor autograd, distributed‑training improvements, new mobile tutorials, performance tools, and several prototype features.
Facebook’s 2020 open‑source report reveals a rapid expansion to over 700 repositories and 1.3 million GitHub followers, highlights AI breakthroughs like the M2M‑100 multilingual model and ReBel algorithm, and showcases frontend tools such as React Native and the newly released Recoil library.
An open‑source PyTorch‑based project enables users to remove unwanted objects from videos simply by drawing a bounding box around them, offering a practical demo, step‑by‑step instructions, and a GitHub repository with over 2 k stars.
This article provides a step‑by‑step illustrated guide to the Transformer architecture, covering its macro structure, detailed self‑attention mechanisms, multi‑head attention, positional encoding, residual connections, decoder operation, training process, loss functions, and includes complete PyTorch and custom Python code examples.
This guide explains five key PyTorch tensor functions—expand, permute, tolist, narrow, and where—detailing their purpose, usage, and code examples so you can manipulate tensors efficiently on CPU or GPU.
This guide walks researchers through selecting a Tencent Cloud GN7 GPU instance, installing NVIDIA drivers, CUDA 10.2, cuDNN, setting up PyTorch and Jupyter, and configuring remote development with PyCharm, enabling efficient, cost‑effective AI development on a Tesla T4 GPU server.
This article introduces Boa, a Node.js library that bridges CPython, enabling JavaScript developers to import and use Python packages—especially machine‑learning libraries like PyTorch—seamlessly within Node.js, complete with practical code examples and API explanations.
This article explains PyTorch's autograd system, covering tensor creation, the requires_grad flag, detaching tensors, disabling gradient tracking with no_grad, the Function class and computational graph, and demonstrates forward and backward passes with code examples illustrating gradient computation and Jacobian‑vector products.
dl_inference is an open‑source inference platform that simplifies deployment of TensorFlow and PyTorch models in production, offering unified gRPC access, load‑balanced multi‑node serving, GPU/CPU options, customizable pre‑ and post‑processing, and extensible architecture for future AI workloads.
A detailed case study reveals why PyTorch training on UFS file storage suffers severe I/O bottlenecks, compares it with local SSD and SSHFS, and presents practical optimizations such as using cv2.imdecode, caching DataLoader handles, and converting small‑file datasets into large UFS files to close the performance gap.
This article explains the concept, mathematics, and practical PyTorch implementation of dilated (or atrous) convolutions, showing how to keep image dimensions unchanged while dramatically increasing the receptive field and discussing their advantages and typical applications.
The article presents a comprehensive overview of 58.com’s AI platform built on Kubernetes, detailing its layered architecture, resource scheduling, offline training pipelines, debugging environment, distributed TensorFlow/PyTorch training, performance benchmarks, and online inference services, highlighting how the system empowers various business units with scalable AI capabilities.
This article introduces ten essential Python libraries—TensorFlow, Scikit‑Learn, NumPy, Keras, PyTorch, LightGBM, Eli5, SciPy, Theano, and Pandas—detailing their features, typical use cases, and adoption in machine‑learning and data‑science projects, while highlighting each library's performance advantages, community support, and integration capabilities to help developers choose the right tool for their AI workflows.
This article provides an in‑depth overview of BERT, covering its NLP background, GLUE benchmark achievements, Transformer‑based architecture, pre‑training strategies (MLM and NSP), downstream fine‑tuning methods, and includes detailed PyTorch code implementations of its core components.
The article explains how transpose (de)convolution works as the spatial inverse of standard convolution, detailing its relationship to fully‑connected layers, padding, stride, output size formulas, odd‑case handling, and practical implementation in frameworks like PyTorch and TensorFlow.
This article reviews major deep‑learning object detection approaches—including one‑stage YOLO and SSD and two‑stage RCNN, Fast RCNN, and Faster RCNN—then provides a step‑by‑step explanation of Faster RCNN’s architecture, region‑proposal network, RoI pooling, loss functions, and sample PyTorch code.
This article introduces five highly rated Python libraries—PyWren, Tfdeploy, Luigi, Kubelib, and PyTorch—that streamline data handling, cloud execution, workflow orchestration, and GPU acceleration, helping machine‑learning engineers boost productivity and tackle complex projects more efficiently.