pip install -UI onepanel
Options:
--version Show the version and exit.
--help Show this message and exit.
Commands:
clone Clone project or dataset from server.
datasets Dataset commands group
download Download a dataset
environments Environment (machine types) commands group
jobs Job commands group
login Login with email and password.
logout Logs the current user out.
machine-types Machine (hardware) commands group
projects Project commands group
pull Pull changes from onepanel (fetch and merge)
push Push changes to onepanel
volume-types Available volume types
workspaces Workspace commands group
LOGIN
after installation, users should log in with valid credentials first via email and password
- after users logs in, user will be able to check available projects with cli
- while logged in users can also create, delete, clone projects / workspaces / datasets under their account with cli
- users can push changes to repository by using 'onepanel push' command.
- users also has access to jobs for parallel training, for more info about jobs please go to JOBS article
LOGOUT
Logs the user out of session
PROJECTS
Projects are the highest level container and are the first thing you will need to create to get started.
https://help.onepanel.io/projects/creating-projects/creating-projects
Commands:
create Create project in new directory.
init Initialize project in current directory.
list Display a list of all projects.
DATASETS
Users can create datasets from local machine using the command <onepanel datasets create 'dataset_name'> or initiate a folder as a dataset by using then running to push in account dataset repository.
Datasets are version controlled containers that can store more than 100Gb of data(can either be images or text files)
- Datasets use AWS and git-lfs to store data, for full guide with AWS please click this link.
https://help.onepanel.io/datasets/creating-datasets
Commands:
create Create dataset in new directory.
init Initialize dataset in current directory.
list Display a list of all datasets.
pull Pull down dataset and any changes
push Push up dataset changes
JOBS
User can create / run jobs with CLI.
Jobs are great for testing different aspects of a model (i.e., training on different versions of a dataset for a model that may include different labeling parameters).
Job command chaining is also available, you can find more information thru this link
Jobs also provides visualizations while training with Tensorboard. For full guide you can refer to this link
Jobs can run in parallel which means:
- users can run training at the same time with different environments, this way users can assess which environment suits the model for better performance
- jobs trained in parallel can have different parameters, this greatly helps hasten users find the best model when training, specially with large datasets.
- this also provide output files while training(if checkpoints are added to code) and after training.
To save your models and any other output, save your data to /onepanel/output.
For full guide please refer to this link
Note: Jobs can only execute code that is committed into the "Code" repository.
https://help.onepanel.io/jobs/creating-jobs
Options:
-m, --machine-type TEXT Machine type ID. Call "onepanel machine-
types list" for IDs. [required]
-e, --environment TEXT Instance template ID. Call "onepanel
environments list" for IDs. [required]
-s, --storage TEXT Storage type ID. [required]
WORKSPACES
Workspaces are version controlled docker instances that can contain source code, libraries & dependencies, and environment settings.
You can also collaborate on workspaces with other users by adding members to your project.
https://help.onepanel.io/workspaces/creating-workspaces
Commands:
create Create a new workspace.
create-from-file Create new workspaces from a json file.
list Show workspaces.
pause Pause the workspace(s) given the uids of the workspace
resume Resume the workspace(s) given the uids of the workspace
terminate Terminate the workspace(s) given the uids of the workspace
ENVIRONMENTS
Below are all the available environment Onepanel can provide, from Tensorflow to H2O, users have a variety of options to use when creating models or scripts for their project, these are all available with both workspaces and jobs.
for information with pricing for types of environment you can refer to this link
Commands:
list Show available environments
Output:
ID ENVIRONMENT
jupyter-py3-tensorflow1.11.0 Python 3, TensorFlow 1.11.0, Jupyter 5.6.0
jupyter-py3-pytorch1.0.0-rc.1 Python 3, PyTorch 1.0, Jupyter 5.6.0
jupyter-py3-tensorflow1.10.0 Python 3, TensorFlow 1.10.0, Jupyter 5.6.0
jupyter-py3-tensorflow1.9.0 Python 3, TensorFlow 1.9.0, Jupyter 5.6.0
jupyter-py3-tensorflow1.8.0 Python 3, TensorFlow 1.8.0, Jupyter 5.5.0
jupyter-py3-pytorch0.4.0 Python 3, PyTorch 0.4.0, Jupyter 5.5.0
jupyter-py3-pytorch0.3.1 Python 3, PyTorch 0.3.1, Jupyter 5.5.0
jupyter-py3-r3.4.3 Python 3, R 3.4.3, Jupyter 5.5.0
h2o3.20.0.1-py3 Python 3, H2O 3.20.0.1
image-annotation-0.2.0 CVAT Annotation Tool, Jupyter 5.5.0
node-red-py3-r3.4.3-node0.18.7 R, Python 3, Node-RED
VOLUME-TYPES
Volume-types provides users information as to how much storage they can use when they run workspaces for their projects, this can either be used for datasets, large data files, images, etc.
for information with pricing for different SSD volumes you can refer to this link
Commands:
list Show available volume types and their IDs
Output:
ID SPECS
default-storage-10 10 GB SSD
default-storage-20 20 GB SSD
default-storage-40 40 GB SSD
default-storage-60 60 GB SSD
default-storage-80 80 GB SSD
default-storage-100 100 GB SSD
default-storage-200 200 GB SSD
default-storage-400 400 GB SSD
default-storage-600 600 GB SSD
default-storage-1000 1 TB SSD
default-storage-2000 2 TB SSD
default-storage-5000 5 TB SSD
default-storage-10000 10 TB SSD
MACHINE-TYPES
Machine type command provides user an overview for available CPU/GPU machines that Onepanel can provide, together with how much they are billed in a hour basis.
Commands:
list Show available machine types
ID SPECS
cpu-2-8 CPU: 2, RAM: 8GB ($0.112/hr)
cpu-8-32 CPU: 8, RAM: 32GB ($0.446/hr)
gpu-4-26-1k80 GPU: 1 (Tesla K80), CPU: 4, RAM: 26GB ($0.750/hr)
gpu-8-52-1v100 GPU: 1 (Tesla V100), CPU: 8, RAM: 52GB ($3.226/hr)