Skip to main content

Abstraction over profile session locations and infrastructure running the analysis.

Project description

xprofiler

The xprofiler SDK and CLI tool provides abstraction over profile session locations and infrastructure running the analysis.

This includes allowing users to create and manage VM instances for TensorBoard instances in regards to profiling workloads for GPU and TPU.

Quickstart

Install Dependencies

xprofiler relies on using gcloud.

The first step is to follow the documentation to install.

Running the initial gcloud setup will ensure things like your default project ID are set.

Create a VM Instance for TensorBoard

To create a TensorBoard instance, you must provide a path to a GCS bucket. It is also useful to define your specific zone.

ZONE=us-central1-a
GCS_PATH="gs://example-bucket/my-profile-data"

xprofiler create -z $ZONE -l $GCS_PATH

When the command completes, you will see it return information about the instance created, similar to below:

Waiting for instance to be created. It can take a few minutes.

Instance for gs://example-bucket/my-profile-data has been created.
You can access it at https://42rc2772e3vg2276-dot-us-central1.notebooks.googleusercontent.com
Instance is hosted at xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef VM.

This will create a VM instance with TensorBoard installed. Note that this initial startup for TensorBoard will take up to a few minutes (typically less than 5 minutes) if you want to connect to the VM's TensorBoard.

List VM Instances

To list the TensorBoard instances created by xprofiler, you can simply run xprofiler list. However, it's recommended to specify the zone (though not required).

ZONE=us-central1-a

xprofiler list -z $ZONE

This will output something like the following if there are instances matching the list criteria:

Log_Directory                                 URL                                                                       Name
-----------------------------------------  ------------------------------------------------------------------------  ------------------------------------------
gs://example-bucket/my-other-profile-data  https://27ac8347d8af0142-dot-us-central1.notebooks.googleusercontent.com  xprof-8187640b-e612-4c47-b4df-59a7fc86b253
gs://example-bucket/my-profile-data        https://42rc2772e3vg2276-dot-us-central1.notebooks.googleusercontent.com  xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef

Note you can specify the GCS bucket to get just that one associated instance:

xprofiler list -l $GCS_PATH

Delete VM Instance

To delete an instance, you'll need to specify either the GCS bucket paths or the VM instances' names. Specifying the zone is required.

# Delete by associated GCS path
xprofiler delete -z $ZONE -l $GCS_PATH

# Delete by VM instance name
VM_NAME="xprof-8187640b-e612-4c47-b4df-59a7fc86b253"
xprofiler delete -z $ZONE --vm-name $VM_NAME

Details on xprofiler

Main Command: xprofiler

The xprofiler command has additional subcommands that can be invoked to create VM instances, list VM instances, delete instances, etc.

However, the main xprofiler command has some additional options without invoking a subcommand.

xprofiler --help

Gives additional information about using the command including flag options and available subcommands. Also can be called with xprofiler -h.

Note that each subcommand has a --help flag that can give information about that specific subcommand. For example: xprofiler list --help

xprofiler --abbrev ...

When invoking a subcommand, typically there is output related to VM instances involved with the subcommand, usually as a detailed table.

In some cases, a user may only want the relevant information (for example a log directory GCS path or VM name instance). This can be particularly useful in scripting with xprofiler by chaining with other commands.

To assist with this, the --abbrev (or equivalent -a) flag will simply print the relevant item (log directory path or VM instance name).

For example, calling xprofiler list might give the following output:

LOG_PATH                                   NAME                                            ZONE
gs://example-bucket/my-other-profile-data  xprof-8187640b-e612-4c47-b4df-59a7fc86b253  us-central1-a
gs://example-bucket/my-profile-data        xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef  us-central1-a

But calling with xprofiler --abbrev list will instead print out an abbreviated form of the above output where each item is displayed on a new line:

xprof-8187640b-e612-4c47-b4df-59a7fc86b253
xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef

Subcommand: xprofiler create

This command is used to create a new VM instance for TensorBoard to run with a given profile log directory GCS path.

Usage details:

xprofiler create
  [--help]
  --log-directory GS_PATH
  [--zone ZONE_NAME]
  [--vm-name VM_NAME]
  [--verbose]

At the successful completion of this command, the information regarding the newly created VM instances is printed out like the example below:

LOG_PATH                            NAME                                            ZONE
gs://example-bucket/my-profile-data xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef  us-central1-a

If the xprofiler abbreviation flag is used, then an abbreviated output is given like so:

xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef

xprofiler create --help

This provides the basic usage guide for the xprofiler create subcommand.

Creating a VM Instance

To create a new VM instance, a user must specify a profile log directory path (a GCS path) as in xprofiler create -l gs://example-bucket/my-profile-data. This will create a VM instance associated with the log directory. The instance will also have TensorBoard installed and setup ready for use.

Note that after the VM creation, it might take a few minutes for the VM instance to fully be ready (installing dependencies, launching TensorBoard, etc.)

It is recommended to also provide a zone with --zone or -z but it is optional.

By default, the VM instance's name will be uniquely created prepended with xprof-. However, this can be specified with the --vm-name or -n flag to give a specific name to the newly created VM.

Lastly, there is a --verbose or -v flag that will provide information as the xprofiler create subcommand runs.

Subcommand: xprofiler list

This command is used to list a VM instances created by the xprofiler tool.

Usage details:

xprofiler list
  [--help]
  [--zone ZONE_NAME]
  [--log-directory GS_PATH [GS_PATH ...]]
  [--filter FILTER_NAME [FILTER_NAME ...]]
  [--verbose]

At the successful completion of this command, the information of matching VM instances is printed out like the example below:

LOG_PATH                                   NAME                                            ZONE
gs://example-bucket/my-other-profile-data  xprof-8187640b-e612-4c47-b4df-59a7fc86b253  us-central1-a
gs://example-bucket/my-profile-data        xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef  us-central1-a

If the xprofiler abbreviation flag is used, then an abbreviated output is given like so:

xprof-8187640b-e612-4c47-b4df-59a7fc86b253
xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef

xprofiler list --help

This provides the basic usage guide for the xprofiler list subcommand.

Listing Specific Subsets

Note that the xprofiler list command will default to listing all VM instances that have the prefix xprof.

However, a specific subset of VM instances can be returned using different options.

Providing Zone

xprofiler list -z $ZONE

Providing the zone is highly recommended since otherwise the command can take a while to search for all relevant VM instances.

Providing GCS Path (Profile Log Directory)

Since xprofiler list is meant to look for VM instances created with xprofiler, it is likely the VM instance of interest is associated with a profile log directory.

To filter for a specific VM instance with an associated log directory, simply use the command like so:

xprofiler list -l $GS_PATH

You can even use multiple log directory paths to find any VMs associated with any of these paths:

xprofiler list -l $GS_PATH_0 $GS_PATH_1 $GS_PATH_2

Subcommand: xprofiler delete

This command is used to delete VM instances, focused on those created by the xprofiler tool.

Usage details:

xprofiler list
  [--help]
  [--log-directory GS_PATH [GS_PATH ...]]
  [--vm-name VM_NAME [VM_NAME ...]]
  [--zone ZONE_NAME]
  [--verbose]

During execution of the delete command, it will prompt the user to confirm each VM to be deleted. (Note the VMs will not be deleted until after confirming or rejecting deletion for each VM.) Output looks similar to the example below:

Found 2 VM(s) to delete.

Log_Directory                              URL                                                                       Name
-----------------------------------------  ------------------------------------------------------------------------  ----------------------------------------------
gs://example-bucket/my-data               https://78700cq71a7f967e-dot-us-central1.notebooks.googleusercontent.com  xprof-8187640b-e612-4c47-b4df-59a7fc86b253
gs://example-bucket/my-data/sub-directory https://7q639a14v4278d50-dot-us-central1.notebooks.googleusercontent.com  xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef

Do you want to continue to delete the VM `xprof-8187640b-e612-4c47-b4df-59a7fc86b253`?
Enter y/n: y

Do you want to continue to delete the VM `xprof-8187640b-e612-4c47-b4df-59a7fc86b253`?
Enter y/n: y

Do you want to continue to delete the VM `xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef`?
Enter y/n: n
Will NOT delete VM `xprof-ev86r7c5-3d09-xb9b-a8e5-a495f5996eef`

xprofiler delete --help

This provides the basic usage guide for the xprofiler delete subcommand.

Providing Zone

xprofiler delete -z $ZONE

Providing the zone is required since otherwise the command can take a while to search for all relevant VM instances that are to be deleted.

Providing GCS Path (Profile Log Directory)

Since xprofiler delete is meant to delete VM instances created with xprofiler, it is likely the VM instance of interest is associated with a profile log directory.

To filter for a specific VM instance with an associated log directory, simply use the command like so:

xprofiler delete -z $ZONE -l $GS_PATH

You can even use multiple log directory paths to find any VMs associated with any of these paths:

xprofiler delete -z $ZONE -l $GS_PATH_0 $GS_PATH_1 $GS_PATH_2

NOTE: The log directory and/or VM name must be specified.

Providing VM Instance Names

Thought the primary method of deletion will likely center around VM instances created with xprofiler, it is sometimes convenient to delete VM instances by their name instead of an associated log directory.

xprofiler delete -z $ZONE --vm-name $VM_NAME $VM_NAME_1

This can also take in multiple VM instances to be deleted:

xprofiler delete -z $ZONE --vm-name $VM_NAME_0 $VM_NAME_1 $VM_NAME_2

Finally, this option can also be used with specifying log directories:

xprofiler delete -z $ZONE -l $GS_PATH --vm-name $VM_NAME_0 $VM_NAME_1 $VM_NAME_2

NOTE: The log directory and/or VM name must be specified.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

cloud_diagnostics_xprof-0.0.9.tar.gz (26.3 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

cloud_diagnostics_xprof-0.0.9-py3-none-any.whl (33.9 kB view details)

Uploaded Python 3

File details

Details for the file cloud_diagnostics_xprof-0.0.9.tar.gz.

File metadata

  • Download URL: cloud_diagnostics_xprof-0.0.9.tar.gz
  • Upload date:
  • Size: 26.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.12.8

File hashes

Hashes for cloud_diagnostics_xprof-0.0.9.tar.gz
Algorithm Hash digest
SHA256 2e1c60cfa5c83784f7f0db082f954ac9b7d117c0cc17e893c4db47fbe7b4126e
MD5 5932abe5f07dc029b74535ebcec66a2e
BLAKE2b-256 85bd11a5b0208c206c3cdb46eb7f8fcc76213f884285826d1cf42f1882f1c6db

See more details on using hashes here.

File details

Details for the file cloud_diagnostics_xprof-0.0.9-py3-none-any.whl.

File metadata

File hashes

Hashes for cloud_diagnostics_xprof-0.0.9-py3-none-any.whl
Algorithm Hash digest
SHA256 170ae1089ded66e5c16e7fdf4e6f0097c1440d27028e2cae40e4eafbaffae156
MD5 eed37d1e677baa83052a8d46947415ec
BLAKE2b-256 cd9abd073cdfbf36edd3fb440f0103762c987820165b7c00ee7b7edfdb99e1ce

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page