grid datastore
Create a low-latency, highly-scalable auto-versioned dataset.
To create a low-latency, auto-versioned dataset, create a datastore
1
grid datastore create --source ./my_dataset --name my_datastore_name
Copied!
Attach a datastore into a Run or Session with:
1
grid run --datastore_name my_datastore_name \
2
--datastore_version 1 \
3
my_script.py
4
--my_data /opt/datastore
Copied!

create

Creates a datastore. Datastores are directories that are compressed into and mounted into your experiment environment.
1
grid datastore create --source path/to/data --name datastore_name
Copied!

--source

Source directory you want to create a datastore from. This directory can contain multiple files and multiple directories.

--name

Name of the dataset. It's name work as a unique ID.

--grid_credential

Cloud credential ID to use.

--compression

Compresses datastores with GZIP when flag is passed.
1
grid datastore create \
2
--source path/to/data \
3
--name datastore_name \
4
--compression
Copied!

delete

Deletes an existing datastore.
1
grid datastore delete --name my_datastore --version x
Copied!

--name

Name of datastore to delete. You will also need to supply a --version.

--version

Version of the datastore to delete. You will also need to supply a --name.

list

Lists all available datastores.
1
grid datastore list
Copied!

resume

Resumes uploading a datastore
grid datastore command lists all available datastores
Last modified 3mo ago