Quick start

This is our quick start guide. Everything you need to know to get started is here.

Installation¶

You can install the lmrtfy package just like any other package:

$ pip install lmrtfy

If you use conda you need to install pip in your conda environment before you can use LMRTFY. This is likely necessary on Windows.

$ conda install pip
$ pip install lmrtfy

Sign-Up¶

To make full use of LMRTFY you need to sign up with us with lmrtfy login.

You can sign up with GitHub to streamline the process.

Calling your first remote function¶

Calling a function that is available in the cloud is as easy as calling a native function.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15

$ ipython
In [1]: from lmrtfy.functions import catalog
# information regarding the configuration
2022-10-12 12:14:28 [36627] INFO {'namespaces': ['examples', 
'<user_namespace>' ]} # (1)!

In [2]: job = catalog.examples.free_fall_lmrtfy(time=100.) # (2)!
2022-10-12 12:16:08 [36627] INFO Job 8GXvA6JREr created. 
Status is ACCEPTED. # (3)!

In [3]: job.ready # (4)!
Out[3]: True

In [4]: job.results  # (5)!
Out[5]: {'velocity': 981.0}

1. If you want to know more about namespaces go here.
2. This calls the function free_fall_lmrtfy in the LMRTFY namespace example. The function is executed on a remote resource that has been provided by us.
3. The job is created and that status is ACCEPTED. Sometimes the computation is so quick that it immediately returns RESULTS_READY.
4. job.ready checks if the results are computed and ready to be fetched with job.results
5. job.results gets the results from LMRTFY as JSON:

   {
     "<variable name>": <value>,
     ...
   }    
   

As you can see calling the function free_fall_lmrtfy looks just like calling any other function in Python; however, it is actually executed on a remote server, in this case on our own server as we provide the example.

Each call to a deployed function returns a Job object if the submission was successful. If an error occurred None is returned. This way we can check if the submission was successful or not.

If you get an error here, please let us know via hello@lmrt.fyi or open an issue.

To get the results from the computation we can simply call job.results. Depending on the size of the results this call will take a while. In the example, this call should at most take a few seconds.

Deploying your first function¶

Deploying your own function is really easy, too.

Let's say you want to calculate how fast a falling object will be after \(x\) seconds.

The code which can be used with LMRTFY is really simple:

1
2
3
4
5
6
7
8
9

from lmrtfy.annotation import variable, result # (1)!

time = variable(200., name="time", min=0, max=1000, unit="s") # (2)!
standard_gravity = 9.81

velocity = standard_gravity*time
print(f"Velocity after {time} seconds is {velocity} m/s.") 

velocity = result(velocity , name="velocity", min=0, max=9810, unit="m") # (3)!

1. variable and result are the imports that are needed to annotate your code to make it work with LMRTFY.
2. variable annotates any inputs of your script.
3. result annotates a result of script. A script can have multiple results.

The highlighted lines have been added or changed to let LMRTFY know about the structure of the script.

Before we can deploy the function we need to run it through the Python interpreter once to create the annotation profile.

$ python free_fall_lmrtfy.py 

To deploy this function on your laptop you simply run

$ lmrtfy deploy free_fall_lmrtfy.py --local

Nobody but you can call the function unless you share it with others.

If no runner is deployed an error will be returned.