Gear Data binding
Binding Content
When creating or restarting a container. You can choose to bind data to a container. Even large-scale datasets can be bound and accessed directly as file systems. The content that can be bound includes the following categories:
- Public dataset
- Personal private dataset
- Working directory for public tutorials
- Personal private work directory for execution
- Upload new code
As shown in the following figure. In the container "Data binding" Interface. All available data for binding will be listed. At the same time, you can also use the name. ID Filter by fields.
Bind to working directory
Every time the container executes a task. They will create a separate storage space and bind it to /hyperai/home catalogue. This is called its working directory. This directory is also Soft link The way points /hyperai/home catalogue. After closing the execution. The contents of the working directory will be saved. This is what we call execution "working directory" .
Please note that. Binding to a working directory actually involves the process of data replication, therefore. The startup time of the container may vary depending on the size of the bound data volume. And this will also occupy the storage space of the working directory.
Bind to data directory
Apart from the working directory. You can also choose to bind data to the following root directory when creating a container:
/hyperai/input/input0/hyperai/input/input1/hyperai/input/input2/hyperai/input/input3/hyperai/input/input4
There are two binding modes for data directories:
- Read and write (read-write)binding: Allow you to add bound data, to update. Delete operation.
- read-only (read-only)binding: You can only read bound data. Unable to add, to update. Delete operation.
Read write binding
For datasets or models with read and write permissions. You can choose "Read write binding" . In this mode. You can directly access the corresponding directory and update the data. The following scenarios are suitable for using read-write binding:
- Preprocess the uploaded raw dataset. Delete unnecessary data.
- Bind two datasets. Extract partial data from a dataset. Then save to another dataset.
- Create an empty dataset version. Then save the data in the container to it.
- Create an empty dataset version. Then use rsync Command to copy local data into it.
Only datasets or models with read-write permissions can be bound for read-write operations. For public datasets without read and write permissions. And other executed work directories. You can only proceed "read-only" binding.
Read only binding
For public datasets without read and write permissions. Public model. And public tutorials created by others, etc. You can only proceed "Read only binding" . In this mode. You can only read data. Unable to add, to update. Delete operation.
As shown in the above figure. For read-only bound data. The container creation page will display corresponding prompt information.