问题
I have a 900mb model and a 9mb model.
I am using the latest version of the Azure ML CLI to register it in my workspace with this command:
az ml model register -n "rj-model" --model-path "models\model_v1.bin -t "model-deployment\model.json"
The 9mb file uploads successfully however the 900 mb file times out with this error:
{'Azure-cli-ml Version': '1.0.60.1', 'Error': AzureHttpError('Operation could not be completed within the specified time. ErrorCode: OperationTimedOut\n<?xml version="1.0" encoding="utf-8"?><Error><Code>OperationTimedOut</Code><Message>Operation could not be completed within the specified time.\nRequestId:2822b750-801e-0061-65fe-71aa88000000\nTime:2019-09-23T11:03:52.5521518Z</Message></Error>',)}
The UI has an even lower timeout threshold.
My thoughts on solving this:
- I need to upload the model to blob storage and access it from the cloud, not locally. I checked the Model Register method and it has an --asset-path property, however this is for an Experiment (I am not running experiments with AML - a Data Scientist is doing them in his own workflow and passing the models over to me for deployment).
- The DataStore looks promising and I could mount my model on a blob (after uploading it with Storage Explorer) and access it - however I can't seem to do this with the CLI. The Python SDK does have a dataset property but not a datastore property and the Register method does not seem to have a cloud option.
So my question is: How can I register large models in Azure ML?
来源：https://stackoverflow.com/questions/58061189/how-to-register-a-local-model-of-900mb-in-azure-machine-learning-service-with-cl