This package allows to run a Jupyter notebook at Google Cloud AI Platform Training Jobs.
Install tool with pip by using pip install gcloud-notebook-training
gcloud-notebook-training [-h] --input-notebook INPUT_NOTEBOOK
[--project-id PROJECT_ID]
[--output-notebook OUTPUT_NOTEBOOK]
[--job-id JOB_ID]
[--region REGION]
[--worker-machine-type WORKER_MACHINE_TYPE]
[--bucket-name BUCKET_NAME]
[--max-running-time MAX_RUNNING_TIME]
[--container-uri CONTAINER_URI]
[--accelerator-type ACCELERATOR_TYPE]
[--service-account SERVICE_ACCOUNT]
The only required parameter is --input-notebook.
This parameter accepts local path or GCS path.
For example:
gcloud-notebook-training --input-notebook /local/path/to/notebook.ipynb
or
gcloud-notebook-training --input-notebook gs://bucket_name/notebook.ipynb
The output-notebook
path can be specified explicitly.
Same as the input-notebook
, this parameter accepts local path or GCS path.
If the output-notebook
is not specified, it will be implied as input-notebook
plus 'output' suffix.
If output-notebook
is a GCS path, then the tool will only submit the training job and exit.
If output-notebook
is a local path, then the tool will wait until the training job succeeds, and then will download the output notebook.
container-uri
parameter specifies the container used by training job.
If this parameter is not specified, the tool will try to pull this information from the notebook metadata.