Operation Process in a Local IDE

ModelArts allows you to remotely access notebook instances from a local IDE to develop AI models based on PyTorch, TensorFlow, or MindSpore. The following figure shows the operation process.

Figure 1 Using a local IDE to develop models

1. Configure a local IDE.

   Configure a local IDE on your PC.

   You can use PyCharm, VS Code, or SSH tools to access a notebook instance from a local IDE. PyCharm and VS Code can be automatically configured using plug-ins or manually configured.

2. Create a notebook instance.

   On the ModelArts management console, create a notebook instance with a proper AI engine and remote SSH enabled.

3. Use the local IDE to remotely access ModelArts DevEnviron.
4. Upload data and code to the development environment.
   • Copy the code to the local IDE, which will automatically synchronize the code to the in-cloud development environment.
   • If the data is less than or equal to 500 MB, directly copy the data to the local IDE.
   • When creating a training job, if the volume of data is greater than 500 MB, upload the data to OBS and then to EVS.
5. Upload the training script and dataset to the OBS directory.
6. Submit a training job.
   • Submit a training job in the local IDE.

     Use ModelArts SDKs. For details, see Creating a Training Job.

   • Submit a training job on the ModelArts management console. For details, see Creating a Training Job.