JarvisLabs Client

Is a Python API for Interacting with Jarvislabs.ai for creating GPU/CPU-powered instances.

Steps to get started

1. Install JLclient
2. Generate API token
3. Authentication
4. Create an Instance object which will let you -
   • Launch GPU/CPU powered instance
   • Pause the instance
   • Resume
   • Destroy
5. Use the User class to
   • get instances that are in running and paused state.
6. Use FileSystem class to manage FileSystem lifecycle.
   • Attach a FileSystem to an instance during creation.
   • Attach a FileSystem to an instance during resume.

Installation

pip install git+https://github.com/jarvislabsai/jlclient.git

Generate API token

you can Generate API token from Jarvislabs.ai.

Imports and configure

from jlclient import jarvisclient
from jlclient.jarvisclient import *
jarvisclient.token = '**************************duWRbO68IiMTkQKWi48'

Managing GPU/CPU powered instances on Jarvislabs.ai

Create Instance

Attributes

Parameter	Type	Description/Values	Default Value
instance_type	str	Choose between GPU or CPU.	GPU
num_gpus / num_cpus	int	According to the instance_type, choose the number of GPUs or CPUs.	1
gpu_type	str	Choose from A100, A5000, A6000, RTX6000Ada, RTX5000.	RTX5000
template	str	Use User.get_templates() in our JLclient to get all templates.	pytorch
script_id	str	If you have a script you can pass it.	None
is_reserved	bool	True refers to an on-demand instance. False refers to a spot instance.	True
duration	str	Choose hour, week, and month. The pricing changes based on the duration..	hour
http_ports	str	As per your requirement, you can specify the ports.	None
storage	int	Choose between 20GB to 2TB.	20

# Launch CPU Instance
instance: Instance = Instance.create('CPU',
                            num_cpus=1,
                            storage=25,
                            name='cpu instance')

# Launch GPU Instance
instance: Instance = Instance.create('GPU',
                            gpu_type='RTX6000Ada',
                            num_gpus=1,
                            storage=50,
                            name='gpu instance')

This should return the Instance object.

you can access the below attributes using the Instance object.

• gpu_type
• num_gpus
• num_cpus
• storage
• name
• machine_id
• script_id
• is_reserved
• duration
• script_args
• http_ports
• template
• url
• endpoints
• ssh_str
• status

Pause the instance

instance.pause()

You can call pause() on any Instance object.

Resume the instance

#Resuming with existing configuration
instance.resume()

#Modifying the parameters like scaling GPUs, switching GPU type, expanding storage, etc.
instance.resume(num_gpus=1,
                gpu_type='RTX5000',
                storage=100)

#Switching from GPU to CPU Instance
instance.resume(num_cpus=1,
                storage=25)

#Switching from CPU to GPU Instance
instance.resume(gpu_type='RTX6000Ada',
                num_gpus=1,
                storage=25)

You can modify an existing instance by changing the below resume parameters.

• num_gpus

• gpu_type

• storage

or just call resume without any parameters to start the instance with the same configuration.

Destroy the instance

instance.destroy()

Pausing inside a Jarvislabs instance

You can use the below function to pause an instance

import os
from jlclient import jarvisclient
from jlclient.jarvisclient import *
jarvisclient.token = "*****"

def pause():
    machine_id = os.getenv('MACHINE_ID')
    instance = User.get_instance(machine_id)
    instance.pause()

User management

The User class comes with the below key functionalities.

• User.get_templates(): Returns the list of templates available.

• User.get_instances(): Returns a list of Instance objects representing all the instances in your account.

• User.get_balance(): Return the balance of the user.

FileSystem

You can use the FileSystem class to create, delete and list all the filesystems in your account.

fs = FileSystem()
fs.list() # List all the filesystems in your account.
fs.create(fs_name='test', storage=500) # Create a new filesystem with the given name and size, returns fs_id that can be used to attach fs to an instance.
fs.delete(fs_id=333) # Delete the filesystem with the given name.

Attach FileSystem to an instance during creation

instance: Instance = Instance.create('GPU',
                            gpu_type='RTX6000Ada',
                            num_gpus=1,
                            storage=50,
                            name='gpu instance',
                            fs_id='332'
                                    )

Attach FileSystem to an instance during resume

instance.resume(fs_id='333')

note

You can get the fs_id by calling fs.list()

Issues/Feature request

Do you like to see any new features, we are all ears. You can drop us an email at hello@jarvislabs.ai or chat with us for requesting any new features or reporting issues.

License

This project is licensed under the terms of the MIT license.