Title: Interface to the Google Cloud Machine Learning Platform
Version: 0.7.1
Description: Interface to the Google Cloud Machine Learning Platform https://cloud.google.com/vertex-ai, which provides cloud tools for training machine learning models.
Depends: R (≥ 3.3.0), tfruns (≥ 1.3)
Imports: config, jsonlite, packrat, processx, rprojroot, rstudioapi, tools, utils, withr, yaml
Suggests: tensorflow (≥ 1.4.2), keras (≥ 2.1.2), knitr, rmarkdown, testthat
License: Apache License 2.0
SystemRequirements: Python (>= 2.7.0)
Encoding: UTF-8
RoxygenNote: 7.3.2
VignetteBuilder: knitr
URL: https://github.com/rstudio/cloudml
BugReports: https://github.com/rstudio/cloudml/issues
NeedsCompilation: no
Packaged: 2025-08-18 22:43:04 UTC; tomasz
Author: Tomasz Kalinowski [cre], Daniel Falbel [aut], Javier Luraschi [aut], JJ Allaire [aut], Kevin Ushey [aut], RStudio [cph]
Maintainer: Tomasz Kalinowski <tomasz@posit.co>
Repository: CRAN
Date/Publication: 2025-08-18 23:50:51 UTC

Interface to the Google Cloud Machine Learning Platform

Description

The cloudml package provides an R interface to Google Cloud Machine Learning Engine, a managed service that enables:

Details

CloudML is a managed service where you pay only for the hardware resources that you use. Prices vary depending on configuration (e.g. CPU vs. GPU vs. multiple GPUs). See https://cloud.google.com/vertex-ai/pricing for additional details.

For documentation on using the R interface to CloudML see the package website at https://github.com/rstudio/cloudml

Author(s)

Maintainer: Tomasz Kalinowski tomasz@posit.co

Authors:

Other contributors:

References

https://github.com/rstudio/cloudml

See Also

Useful links:

Deploy SavedModel to CloudML

Description

Deploys a SavedModel to CloudML model for online predictions.

Usage

cloudml_deploy(
  export_dir_base,
  name,
  version = paste0(name, "_1"),
  region = NULL,
  config = NULL
)

Arguments

export_dir_base

A string containing a directory containing an exported SavedModels. Consider using tensorflow::export_savedmodel() to export this SavedModel.

name

The name for this model (required)

version

The version for this model. Versions start with a letter and contain only letters, numbers and underscores. Defaults to name_1

region

The region to be used to deploy this model.

config

A list, YAML or JSON configuration file as described https://cloud.google.com/vertex-ai.

See Also

cloudml_predict()

Other CloudML functions: cloudml_predict(), cloudml_train()

Perform Prediction over a CloudML Model.

Description

Perform online prediction over a CloudML model, usually, created using cloudml_deploy()

Usage

cloudml_predict(instances, name, version = paste0(name, "_1"), verbose = FALSE)

Arguments

instances

A list of instances to be predicted. While predicting a single instance, list wrapping this single instance is still expected.

name

The name for this model (required)

version

The version for this model. Versions start with a letter and contain only letters, numbers and underscores. Defaults to name_1

verbose

Should additional information be reported?

See Also

cloudml_deploy()

Other CloudML functions: cloudml_deploy(), cloudml_train()

Train a model using Cloud ML

Description

Upload a TensorFlow application to Google Cloud, and use that application to train a model.

Usage

cloudml_train(
  file = "train.R",
  master_type = NULL,
  flags = NULL,
  region = NULL,
  config = NULL,
  collect = "ask",
  dry_run = FALSE
)

Arguments

file

File to be used as entrypoint for training.

master_type

Training master node machine type. "standard" provides a basic machine configuration suitable for training simple models with small to moderate datasets. See the documentation at https://cloud.google.com/vertex-ai/docs/reference/rest/v1/MachineSpec for details on available machine types.

flags

Named list with flag values (see flags()) or path to YAML file containing flag values.

region

The region to be used for training.

config

A list, YAML or JSON configuration file as described https://cloud.google.com/vertex-ai.

collect

Logical. If TRUE, collect job when training is completed (blocks waiting for the job to complete). The default ("ask") will interactively prompt the user whether to collect the results or not.

dry_run

Triggers a local dry run over the deployment phase to validate packages and packing work as expected.

See Also

job_status(), job_collect(), job_cancel()

Other CloudML functions: cloudml_deploy(), cloudml_predict()

Examples

## Not run: 
library(cloudml)

gcloud_install()
job <- cloudml_train("train.R")

## End(Not run)

Executes a Google Cloud Command

Description

Executes a Google Cloud command with the given parameters.

Usage

gcloud_exec(..., args = NULL, echo = TRUE, dry_run = FALSE)

Arguments

...

Parameters to use specified based on position.

args

Parameters to use specified as a list.

echo

Echo command output to console.

dry_run

Echo but not execute the command?

Examples

## Not run: 
gcloud_exec("help", "info")

## End(Not run)

Initialize the Google Cloud SDK

Description

Initialize the Google Cloud SDK

Usage

gcloud_init()

See Also

Other Google Cloud SDK functions: gcloud_install(), gcloud_terminal()

Install the Google Cloud SDK

Description

Installs the Google Cloud SDK which enables CloudML operations.

Usage

gcloud_install(update = TRUE)

Arguments

update

Attempt to update an existing installation.

See Also

Other Google Cloud SDK functions: gcloud_init(), gcloud_terminal()

Examples

## Not run: 
library(cloudml)
gcloud_install()

## End(Not run)

Create an RStudio terminal with access to the Google Cloud SDK

Description

Create an RStudio terminal with access to the Google Cloud SDK

Usage

gcloud_terminal(command = NULL, clear = FALSE)

Arguments

command

Command to send to terminal

clear

Clear terminal buffer

Value

Terminal id (invisibly)

See Also

Other Google Cloud SDK functions: gcloud_init(), gcloud_install()

Gcloud version

Description

Get version of Google Cloud SDK components.

Usage

gcloud_version()

Value

a list with the version of each component.

Copy files to / from Google Storage

Description

Use the ⁠gsutil cp⁠ command to copy data between your local file system and the cloud, copy data within the cloud, and copy data between cloud storage providers.

Usage

gs_copy(source, destination, recursive = FALSE, echo = TRUE)

Arguments

source

The file to be copied. This can be either a path on the local filesystem, or a Google Storage URI (e.g. ⁠gs://[BUCKET_NAME]/[FILENAME.CSV]⁠).

destination

The location where the source file should be copied to. This can be either a path on the local filesystem, or a Google Storage URI (e.g. ⁠gs://[BUCKET_NAME]/[FILENAME.CSV]⁠).

recursive

Boolean; perform a recursive copy? This must be specified if you intend on copying directories.

echo

Echo command output to console.

Google storage bucket path that syncs to local storage when not running on CloudML.

Description

Refer to data within a Google Storage bucket. When running on CloudML the bucket will be read from directly. Otherwise, the bucket will be automatically synchronized to a local directory.

Usage

gs_data_dir(url, local_dir = "gs", force_sync = FALSE, echo = TRUE)

Arguments

url

Google Storage bucket URL (e.g. ⁠gs://<your-bucket>⁠).

local_dir

Local directory to synchonize Google Storage bucket(s) to.

force_sync

Force local synchonization even if the data directory already exists.

echo

Echo command output to console.

Details

This function is suitable for use in TensorFlow APIs that accept gs:// URLs (e.g. TensorFlow datasets). However, many package functions accept only local filesystem paths as input (rather than gs:// URLs). For these cases you can the gs_data_dir_local() function, which will always synchronize gs:// buckets to the local filesystem and provide a local path interface to their contents.

Value

Path to contents of data directory.

See Also

gs_data_dir_local()

Get a local path to the contents of Google Storage bucket

Description

Provides a local filesystem interface to Google Storage buckets. Many package functions accept only local filesystem paths as input (rather than gs:// URLs). For these cases the gcloud_path() function will synchronize gs:// buckets to the local filesystem and provide a local path interface to their contents.

Usage

gs_data_dir_local(url, local_dir = "gs", echo = FALSE)

Arguments

url

Google Storage bucket URL (e.g. ⁠gs://<your-bucket>⁠).

local_dir

Local directory to synchonize Google Storage bucket(s) to.

echo

Echo command output to console.

Details

If you pass a local path as the url it will be returned unmodified. This allows you to for example use a training flag for the location of data which points to a local directory during development and a Google Cloud bucket during cloud training.

Value

Local path to contents of bucket.

Note

For APIs that accept gs:// URLs directly (e.g. TensorFlow datasets) you should use the gs_data_dir() function.

See Also

gs_data_dir()

Alias to gs_data_dir_local() function

Description

This function is deprecated, please use gs_data_dir_local() instead.

Usage

gs_local_dir(url, local_dir = "gs", echo = FALSE)

Arguments

url

Google Storage bucket URL (e.g. ⁠gs://<your-bucket>⁠).

local_dir

Local directory to synchonize Google Storage bucket(s) to.

echo

Echo command output to console.

See Also

gs_data_dir_local()

Synchronize content of two buckets/directories

Description

The gs_rsync function makes the contents under destination the same as the contents under source, by copying any missing files/objects (or those whose data has changed), and (if the delete option is specified) deleting any extra files/objects. source must specify a directory, bucket, or bucket subdirectory.

Usage

gs_rsync(
  source,
  destination,
  delete = FALSE,
  recursive = FALSE,
  parallel = TRUE,
  dry_run = FALSE,
  options = NULL,
  echo = TRUE
)

Arguments

source

The file to be copied. This can be either a path on the local filesystem, or a Google Storage URI (e.g. ⁠gs://[BUCKET_NAME]/[FILENAME.CSV]⁠).

destination

The location where the source file should be copied to. This can be either a path on the local filesystem, or a Google Storage URI (e.g. ⁠gs://[BUCKET_NAME]/[FILENAME.CSV]⁠).

delete

Delete extra files under destination not found under source By default extra files are not deleted.

recursive

Causes directories, buckets, and bucket subdirectories to be synchronized recursively. If you neglect to use this option gs_rsync() will make only the top-level directory in the source and destination URLs match, skipping any sub-directories.

parallel

Causes synchronization to run in parallel. This can significantly improve performance if you are performing operations on a large number of files over a reasonably fast network connection.

dry_run

Causes rsync to run in "dry run" mode, i.e., just outputting what would be copied or deleted without actually doing any copying/deleting.

options

Character vector of additional command line options to the gsutil rsync command (as specified at https://cloud.google.com/storage/docs/gsutil/commands/rsync).

echo

Echo command output to console.

Executes a Google Utils Command

Description

Executes a Google Utils command with the given parameters.

Usage

gsutil_exec(..., args = NULL, echo = FALSE)

Arguments

...

Parameters to use specified based on position.

args

Parameters to use specified as a list.

echo

Echo command output to console.

Cancel a job

Description

Cancel a job.

Usage

job_cancel(job = "latest")

Arguments

job

Job name or job object. Pass "latest" to indicate the most recently submitted job.

See Also

Other job management functions: job_collect(), job_list(), job_status(), job_stream_logs(), job_trials()

Collect job output

Description

Collect the job outputs (e.g. fitted model) from a job. If the job has not yet finished running, job_collect() will block and wait until the job has finished.

Usage

job_collect(
  job = "latest",
  trials = "best",
  destination = "runs",
  timeout = NULL,
  view = interactive()
)

Arguments

job

Job name or job object. Pass "latest" to indicate the most recently submitted job.

trials

Under hyperparameter tuning, specifies which trials to download. Use "best" to download best trial, "all" to download all, or a vector of trials c(1,2) or 1.

destination

The destination directory in which model outputs should be downloaded. Defaults to runs.

timeout

Give up collecting job after the specified minutes.

view

View the job results after collecting it. You can also pass "save" to save a copy of the run report at tfruns.d/view.html

See Also

Other job management functions: job_cancel(), job_list(), job_status(), job_stream_logs(), job_trials()

List all jobs

Description

List existing Google Cloud ML jobs.

Usage

job_list(
  filter = NULL,
  limit = NULL,
  page_size = NULL,
  sort_by = NULL,
  uri = FALSE
)

Arguments

filter

Filter the set of jobs to be returned.

limit

The maximum number of resources to list. By default, all jobs will be listed.

page_size

Some services group resource list output into pages. This flag specifies the maximum number of resources per page. The default is determined by the service if it supports paging, otherwise it is unlimited (no paging).

sort_by

A comma-separated list of resource field key names to sort by. The default order is ascending. Prefix a field with ~ for descending order on that field.

uri

Print a list of resource URIs instead of the default output.

See Also

Other job management functions: job_cancel(), job_collect(), job_status(), job_stream_logs(), job_trials()

Current status of a job

Description

Get the status of a job, as an R list.

Usage

job_status(job = "latest")

Arguments

job

Job name or job object. Pass "latest" to indicate the most recently submitted job.

See Also

Other job management functions: job_cancel(), job_collect(), job_list(), job_stream_logs(), job_trials()

Show job log stream

Description

Show logs from a running Cloud ML Engine job.

Usage

job_stream_logs(
  job = "latest",
  polling_interval = getOption("cloudml.stream_logs.polling", 5),
  task_name = NULL,
  allow_multiline_logs = FALSE
)

Arguments

job

Job name or job object. Pass "latest" to indicate the most recently submitted job.

polling_interval

Number of seconds to wait between efforts to fetch the latest log messages.

task_name

If set, display only the logs for this particular task.

allow_multiline_logs

Output multiline log messages as single records.

See Also

Other job management functions: job_cancel(), job_collect(), job_list(), job_status(), job_trials()

Current trials of a job

Description

Get the hyperparameter trials for job, as an R data frame

Usage

job_trials(x)

Arguments

x

Job name or job object.

See Also

Other job management functions: job_cancel(), job_collect(), job_list(), job_status(), job_stream_logs()