GPUDashboard

A simple dashboard for NVIDIA GPU

Demo

Example

Requirement

• Python 2.7 or 3.6
• NVIDIA-sim
• A Firebase realtime database
• Linux-like OS

Setup

1. Create a Firebase Realtime database

2. Set the rules to

{
  "rules": {
    ".read": true,
    ".write": true
  }
}

3. Go to Project overview click Add Firebase to your web app and copy following part.

  var config = {
    apiKey: "XXXXXXXXXXXXXXXXXXXXXXXXXXXX",
    authDomain: "XXXXX.firebaseapp.com",
    databaseURL: "https://XXXXXX.firebaseio.com",
    projectId: "XXXXXXX",
    storageBucket: "XXXXXXX.appspot.com",
    messagingSenderId: "XXXXXXXXXXX"
  };

4. On the servers that have NVIDIA GPU(s) installed.

pip install GPUDashboard
GPUDashboard -n your_server_name -i 20 -u your_databaseURL > GPUDashboard.log 

# your_server_name is the name you want to give your server e.g. MyFirstServer
# -i is the interval of GPU information updating
# your_databaseURL is the databaseURL obtained froom Firebase as shown above

Now, the server GPU information is post to the firebase. *If you have many servers, all of them can make use of the same database you created in Firebase. You only need to specify different names for "your_server_name" when you start the GPUDashboard in the command line on the different servers.

5. Download ViewStatus.html and open with text editor then replace the "config".

<html>
    <header>
      <script>
        var config = {
            apiKey: "XXXXXXXXXXXXXXXXXXXXXXXXXXXX",
            authDomain: "XXXXX.firebaseapp.com",
            databaseURL: "https://XXXXXX.firebaseio.com",
            projectId: "XXXXXXX",
            storageBucket: "XXXXXXX.appspot.com",
            messagingSenderId: "XXXXXXXXXXX"
          };
      </script>
      <link rel="stylesheet" href="https://fonts.googleapis.com/icon?family=Material+Icons"/>

6. Open the "modified ViewStatus.html" with browser.