Prepare
Pycharm requires a professional version and an SSH connection tool is required.
AutoDL registers and creates a container instance.
principle
Pycharm edits the code, uses SSH tools to connect to the cloud server to upload the code, and the cloud server uses the uploaded code and the data set and environment on the existing server for deep learning.
In general, the code is edited and uploaded locally, while the cloud server runs the code.
Therefore, the following three problems need to be solved:
1. Cloud service environment creation
2. Local code upload
3. Data set upload on the cloud server
Cloud service environment creation
An environment can directly select an image when creating a container instance.
To get started we can use the base image, which has some basic packages.
After using it for a period of time, that is, after installing some packages, you can also package the image, customize your own image and then migrate it. The detailed operation will not be repeated.
code upload
AutoDL creates an instance
Login command format = ssh -p Port Username@Host
Select the compiler Interpreter in the remote server, and configure the synchronization mapping of the folder Sync folders
. The result of file synchronization: the files in the local project will be automatically synchronized to /root/autodl-tmp/remotetorch in the remote server. You can also manually
select the file Upload, right-click the file to be uploaded and select Upload to xxx agent in Deployment
Dataset upload on cloud server
1. Server internal data set
You can copy and paste directly into your own project folder
2. Online disk authorization upload
After clicking download, the download path will be prompted
3. Xftp upload
Download the data set locally from the network, and then use Xshell to upload the data set to the server
Effect
other
1. Change name
Modifying the unified name can ensure that there will be no confusion in the case of multiple remote servers.
2. Empty the Trash
sudo rm -rf .local/share/Trash/*