Using Command Line Tools
Brief introduction
We implemented the Cloud ML command line tools based on the Python SDK, enabling easy access to the Xiaomi Cloud ML service.
Initialize Configuration File
cloudml init
View help
cloudml -h
TrainJob Related Commands
List Training Jobs
cloudml jobs list
Submit training jobs
cloudml jobs submit -n linear1 -m trainer.task -u fds://cloud-ml-test/linear/trainer-1.0.tar.gz
Command line parameters:
Required | Argument | Type | Example |
---|---|---|---|
Yes | -n --name | string | linear1 |
Yes | -m --module_name | string | trainer.task |
Yes | -u --trainer_uri | string | fds://cloud-ml-test/linear/trainer-1.0.tar.gz |
No | -f --filename | string | tensorflow.json {"job_name": "linear1", "module_name": "trainer.task", "trainer_uri": "fds://cloud-ml-test/linear/trainer-1.0.tar.gz"} |
No | -a --job_args | string | "--max_epochs 10 --optimizer sgd" |
No | -c --cpu_limit | int | 0.1 |
No | -M --memory_limit | string | 100M |
No | -g --gpu_limit | int | 1 |
No | -p --ps_count | int | 1 |
No | -w --worker_count | int | 1 |
No | -d --docker_image | string | cr.d.xiaomi.net/cloud-ml/train-tensorflow-cpu:0.11.0-xm1.0.0 |
No | -F --framework | string | tensorflow |
No | -V --framework_version | string | 0.11.0-xm1.0.0 |
View training job
cloudml jobs describe linear1
View training job log
cloudml jobs logs linear1
Delete Training Job
cloudml jobs delete linear1
View job creation log
cloudml jobs events linear1
Other commands
List Quotas
cloudml quota list
List frameworks
cloudml framework list
View Versions
cloudml -v