Rapidly Building Graph Algorithm Tasks on Tencent Cloud TI-ONE Platform

The Intelligent Titanium (TI‑ONE) machine learning platform is a one‑stop service for AI engineers, offering end‑to‑end support from data preprocessing to model training, evaluation, and deployment, with a rich library of algorithm components and frameworks.

TI‑ONE addresses common challenges faced by AI practitioners, including limited GPU resources, rapidly evolving frameworks, high learning curves for machine learning and deep learning, time‑consuming hyper‑parameter tuning, and the need for fast iteration in complex business scenarios.

To solve these problems, TI‑ONE provides on‑demand compute resources, a visual drag‑and‑drop task designer, integration of popular frameworks such as PyTorch, TensorFlow, PySpark, and Angel, built‑in algorithm libraries (CNN, RNN, clustering, visualization, etc.), flexible execution modes, one‑click deployment, and an interactive Notebook environment.

The platform also offers open datasets and a visual modeling canvas where users can select algorithm modules, connect them on a flowchart, configure parameters via a side panel, and run the workflow to obtain logs and intermediate results.

TI‑ONE includes two Angel algorithm options: the Spark‑on‑Angel framework for custom code execution and pre‑packaged Angel algorithm components (graph, PyTONA, machine‑learning algorithms). Detailed usage documentation and parameter descriptions are provided for each component.

For custom code training, users drag the Spark‑on‑Angel component onto the canvas, fill in job JAR, main class, and program arguments, and select resource specifications (CPU instance types, executor/driver counts). The platform simplifies submission, provides unified log access, and supports resource auto‑release.

As a concrete example, the article demonstrates building a GraphSage graph algorithm task: the user drags the GraphSage module and a COS dataset onto the canvas, configures IO paths and algorithm parameters (batch size, learning rate, partitions, epochs, etc.), and runs the job. After execution, logs and model artifacts are available via the platform.

In summary, from data upload to algorithm selection, parameter configuration, workflow construction, log management, model building, and result visualization, the entire AI development lifecycle can be performed within TI‑ONE.