Here you will find answers to the most frequently asked questions.
What environment (SAPI) should I run Rubix ML in?#
All Rubix ML projects are designed to run from the PHP command line interface (CLI). The reason almost always boils down to performance and memory consumption.
If you would like to serve your models in production, the preferred method is to use the Server library to spin up a high-performance standalone model server from the command line. If you plan to implement your own model server, we recommend using an asynchronous event loop such as React PHP or Swoole to prevent the model from having to be loaded into memory on each request.
To run a PHP script using the command line interface (CLI), open a terminal window and enter:
$ php example.php
Note: The PHP interpreter must be installed and in your default PATH for the above syntax to work.
I'm getting out of memory errors.#
Try adjusting the
memory_limit option in your php.ini file to something more reasonable. We recommend setting this to
-1 (no limit) or slightly below your device's memory supply for best results.
You can temporarily set the
memory_limit in your script by using the
Note: Training can require a lot of memory. The amount necessary will depend on the amount of training data and the size of your model. If you have more data than you can hold in memory, some learners allow you to train in batches. See the section on Training for more information.
Training is slower than usual.#
Training time depends on a number of factors including size of the dataset and complexity of the model. If you believe that training is taking unusually long then check the following factors.
- Xdebug or other debuggers are not enabled.
- You have enough RAM to hold the dataset and model in memory without swapping to disk.
What is a Tuple?#
A tuple is a way to denote an immutable sequential heterogeneous list with a predefined length. An n-tuple is a tuple with the length of n. In some languages, tuples are a separate data type and their properties such as immutability are enforced by the compiler/interpreter. In PHP, tuples are denoted by sequential arrays which are mutable as a side effect.
$tuple = ['first', 'second', 0.001]; // a 3-tuple
Does Rubix ML support multiprocessing/multithreading?#
Yes, learners that support parallel processing (multiprocessing or multithreading) do so by utilizing a pluggable parallel computing backend or extension under the hood.
Does Rubix ML support Deep Learning?#
Does Rubix ML support Reinforcement Learning?#
Not currently. Rubix ML is for supervised and unsupervised learning only.
Does Rubix ML support time series data?#
Yes and no. Currently, the library treats time series data like any other continuous feature. In the future, we may add algorithms that work specifically with a separate time component.
How can I contribute to the project?#
Anyone is welcome to contribute to Rubix ML. See the CONTRIBUTING guide in the project root for more info.
How can I become a sponsor?#
Check out our funding sources here and consider donating to the cause.