blob: db2b843a2ca817df772cded34e9d38f783b7eee8 (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
|
## Compatibility
This Feature adds shared libraries for NVIDIA CUDA and is only useful for devcontainers that run on a host machine with an NVIDIA GPU. Within your devcontainer, use the `nvidia-smi` command to ensure that your GPU is available for CUDA.
If the `nvidia-smi` command is not available within your devcontainer, you may need to complete the following steps:
### Install the NVIDIA Container Toolkit
Follow [NVIDIA's instructions to install the NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/overview.html) on your host machine. The NVIDIA Container Toolkit is available on a variety of Linux distributions. Make sure you have installed the NVIDIA driver for your Linux distribution before installing the NVIDIA Container Toolkit.
### Enable GPU passthrough
Enable GPU passthrough to your devcontainer by using `hostRequirements`. Here's an example of a devcontainer with this property:
```json
{
"hostRequirements": {
"gpu": "optional"
}
}
```
> Note: Setting `gpu` property's value to `true` will work with GPU machine types, but fail with CPUs. Hence, setting it to `optional` works in both cases. See [schema](https://containers.dev/implementors/json_schema/#base-schema) for more configuration details.
## OS Support
This Feature should work on recent versions of Debian/Ubuntu-based distributions with the `apt` package manager installed.
`bash` is required to execute the `install.sh` script.
|
