Skip to main content

Manage Your Endpoint

Run Anywhere

Run in CLI

fedml model run -e $endpoint_id $json_string

For more information, please refer to FedML Model Run CLI.

Other Language Support

We support language bindings for Python, Java, JQuery, etc. After you deploy a model card to Nexus AI Platform, you can go to "Deploy" -> "Endpoints" -> Your Endpoint.

runAnywhere.jpg

Update Your Endpoint

Update the model card

Suppose you make some changes to your model card, from:

def predict(self, request):
return {f"AlohaV1From{self.worker_id}": request}

to:

def predict(self, request):
return {f"AlohaV2From{self.worker_id}": request}

Recreate the model card, then upload to the Nexus AI Platform. By using:

fedml model create -n $model_card_name -cf $config_file
fedml model push -n $model_card_name

You can see the model card has been updated on the Nexus AI Platform. checkNewVersion.png

Edit the endpoint

Assume you have a existed endpoint under "Deploy" -> "Endpoints" -> Your Endpoint. ExistedEp.png

Click "Edit" button. EditButton.png

Select the new model card version. EditPage.png

click "Update" button. UpdateButton.png

By default, the update will use a scrolling update strategy. Which means, it only updates a small portion of the devices at a time. This can ensure that this endpoint is always available when updating.

You can change the ratio of the devices that you want to update at a time, using the following parameter in the config.yaml file:

max_unavailable_rate: 0.5

rollingUpdate.png