forked from google-ai-edge/model-explorer
-
Notifications
You must be signed in to change notification settings - Fork 3
Graph Execution
Marco Campos edited this page Oct 28, 2024
·
1 revision
If the graph has no errors, a button will be available to execute the graph, that sends the execute command to the server.

The UI will then send graph information to the server so it can compile and run the graph again. Here are the types for the request:
export declare interface AdapterExecuteCommand extends ExtensionCommand {
cmdId: 'execute';
modelPath: string;
settings: Record<string, any>;
deleteAfterConversion: boolean;
}So far, the settings property doesn't have any defined information to send. This can be extended later.
The server response should follow the type below:
export declare interface AdapterExecuteResponse extends ExecutionCommand {
error?: string;
log_file: string;
stdout: string;
perf_data?: NodeDataProviderData
}In case the error property is present, the UI will show an error popup to the user displaying what errors occured.
If the perf_data property is present, if should follow the Cusom User Data format from Google's documentation. This will auto apply to the graph.