BeakerUtil
This project is a collection of command-line utilities for Beaker within AllenAI.
Installation
- Ensure that Beaker is set up locally by following these instructions.
- Install this project with
pip install beakerutil.
[!NOTE] Instead of installing
BeakerUtil, you can useuvx beakerutilto directly run the tool. On my machine, I set up the aliasbu="uvx beakerutil"for convenience.
Usage
The main script is beakerutil, which is the entrypoint for all utilities, which are specified as subcommands.
Run beakerutil -h for more information.
Using beakerutil launch
With BeakerUtil, you can specify different launch configurations in ~/.beakerutil/launch.conf with different parameters, which can effectively be used as an alias to launch sessions quickly. For example, with my config below, I can run beakerlaunch phobos to quickly launch a CPU-only job on phobos for managing files on WEKA.
Clusters can be specified with a regex to match to multiple clusters. Additionally, the DEFAULT configuration specify parameters to be applied for every launch configuration, but can be overridden. For reference, this is my launch.conf:
DEFAULT:
budget: ai2/prior
workspace: ai2/abhayd
env_secrets:
AWS_SECRET_ACCESS_KEY: AWS_ACCESS_KEY
AWS_ACCESS_KEY_ID: AWS_ACCESS_KEY_ID
WANDB_API_KEY: WANDB_API_KEY
HF_TOKEN: HF_TOKEN
GEMINI_API_KEY: GEMINI_API_KEY
OPENAI_API_KEY: OPENAI_API_KEY
DOCKER_PAT: DOCKER_PAT
phobos:
cluster: ai2/phobos-cirrascale
mounts:
- src: weka
ref: prior-default
dst: /weka/prior
- src: weka
ref: oe-training-default
dst: /weka/oe-training-default
gpu:
cluster: ".*-cirrascale.*"
gpus: 1
mounts:
- src: weka
ref: prior-default
dst: /weka/prior
- src: weka
ref: oe-training-default
dst: /weka/oe-training-default
elanding:
cluster: "ai2/prior-elanding.*"
Extra Arguments
Extra arguments to pass to beaker session create can also be passed as additional positional arguments preceded by the delimeter --. For example, to name the interactive session, you could do:
beakerlaunch phobos -- -n foo
Using beakerutil monitor
beakerutil monitor can be used to monitor the resource usage of your running experiments. Example output could look like this:
07/17/2025 13:05:59
+----------------------------+---------------------------------+---------+---------------------+-----------------------+-------------------+---------+--------------------+---------------------+
| Job | Hostname | CPU % | RAM | GPU(s) | VRAM | GPU % | Network (In/Out) | Disk (Write/Read) |
+============================+=================================+=========+=====================+=======================+===================+=========+====================+=====================+
| 01K0CZTBC3AFQ91945TFSFH80G | jupiter-cs-aus-145.reviz.ai2.in | 2.90% | 793.5MiB / 1.968TiB | NVIDIA H100 80GB HBM3 | 1 MiB / 81559 MiB | 0 % | 41.7MB / 634kB | 3.92MB / 60.4MB |
+----------------------------+---------------------------------+---------+---------------------+-----------------------+-------------------+---------+--------------------+---------------------+
| 01K0CSRRM61YE677QB3BG1CAY4 | neptune-cs-aus-265.reviz.ai2.in | 542.51% | 14.83GiB / 1008GiB | | | | 3.57GB / 3.59MB | 183MB / 14.5GB |
+----------------------------+---------------------------------+---------+---------------------+-----------------------+-------------------+---------+--------------------+---------------------+
By default, it monitors continuously, updating every 2 seconds. Use the -n flag to modify the update interval, or use --once to print the usage once before exiting.
Using beakerutil list
beakerutil list is a straightforward command, it takes no arguments and prints the currently running beaker jobs, both interactive and noninteractive. For example, the output may look like this:
Interactive sessions:
0: Session 01JBWFVSS0HNZWW5CT8C84F0J1 using 1 GPU(s) on prior-elanding-62.reviz.ai2.in, status=idle
In this case, there is only one session, with index 0 and ID 01JBWFVSS0HNZWW5CT8C84F0J1.
Using beakerutil attach
beakerutil attach is a utility for connecting to an existing session with beaker session attach. For example, say the output of beakerutil list looks like this:
Interactive sessions:
0: Session 01JBWFVSS0HNZWW5CT8C84F0J1 using 1 GPU(s) on prior-elanding-62.reviz.ai2.in, status=idle
Running beakerutil attach 0 will connect to the session at index 0. Alternatively, you can specify the session ID or name with the --id or --name flags, respectively. If you only have one session running (as is the case here), you can simply run beakerutil attach (with no arguments) to attach to that session.
Using beakerutil stop
beakerutil stop can be used to easily cancel jobs or sessions. It has a similar usage as beakerutil attach, simply specify an index, ID, or name.
Shortcuts
Some shorthand commands are provided for convenience:
beakerlaunchis a shorthand forbeakerutil launch. All arguments are forwarded.
Metadata
Release files for BeakerUtil 0.3.10
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| beakerutil-0.3.10.tar.gz | 11.3 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| beakerutil-0.3.10-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 23.8 kB
Release files / beakerutil-0.3.10.tar.gz
| Download URL | beakerutil-0.3.10.tar.gz |
|---|---|
| Size | 11.3 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
56729de68118f081f644358be24096ce50088fd0ecaabd62f9e3871ce6b78386
|
|
BLAKE2b-256 checksum How to use checksums |
10df33a8f5117443f7ae298d838ac1258149589e849a7c34cd50cd31516727b3
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Dec 8, 2025.
Transparency logRelease files / beakerutil-0.3.10-py3-none-any.whl
| Download URL | beakerutil-0.3.10-py3-none-any.whl |
|---|---|
| Size | 12.5 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
a00c592c61245c24e4aac7861187fdbe87555746db0c3a4c778bdbc6fd2ecd7c
|
|
BLAKE2b-256 checksum How to use checksums |
8be63e133278fbc07976b7736fd11a600a52667ee2996bda8aaf921c5dfea852
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Dec 8, 2025.
Transparency log