modelarts-umn/umn/source/training_management/introduction_to_model_training.rst

original_name

modelarts_23_0044.html

Introduction to Model Training

ModelArts provides model training for you to view the training effect, based on which you can adjust your model parameters. You can select resource pools (CPU or GPU) with different instance flavors for model training. In addition to the models developed by users, ModelArts also provides built-in algorithms. You can directly adjust parameters of the built-in algorithms, instead of developing a model by yourself, to obtain a satisfactory model.

Description of the Model Training Function

Table 1 Function description

Function	Description	Reference
Built-in algorithms	Based on the frequently-used AI engines in the industry, ModelArts provides built-in algorithms to meet a wide range of your requirements. You can directly select the algorithms for training jobs, without concerning model development.	Built-in Algorithms <modelarts_23_0156>
Training job management	You can create training jobs, manage training job versions, and view details of training jobs, and evaluation details.	Introduction to Training Jobs <modelarts_23_0046> Managing Training Job Versions <modelarts_23_0047> Viewing Job Details <modelarts_23_0048>
Job parameter management	You can save the parameter settings of a training job (including the data source, algorithm source, running parameters, resource pool parameters, and more) as a job parameter, which can be directly used when you create a training job, eliminating the need to set parameters one by one. As such, the configuration efficiency can be greatly improved.	Managing Job Parameters <modelarts_23_0049>
Model training visualization	TensorBoard and MindInsight effectively display the computational graph of a model in the running process, the trend of all metrics in time, and the data used in the training.	Managing Visualization Jobs <modelarts_23_0050>