Getting Started

This document will show you how to get up and running with TensorCI. You will learn how to create your TensorCI account, install the CLI, and create your first TensorCI project.

Create an Account

Before doing anything else, you will need to create a TensorCI account. Doing this is easy – just visit the TensorCI home page and sign up with GitHub.

Once you’ve signed in, you should be taken to the TensorCI dashboard and prompted to create a basic auth password. This password is required in order to login to the TensorCI CLI. If you’re not automatically prompted to do this on your first visit to the dashboard, you can use this link instead.

Install the CLI

The TensorCI CLI is the easiest way to interact with your TensorCI resources. It provides commands for creating datasets, deploying your models for training, downloading your trained models, and serving model predictions from your TensorCI API, among others.

Assuming you have Python already, you can install the CLI with pip:

$ pip install tensorci

Login from the CLI

Now that you have the tensorci command-line tool, you should be able to login to your TensorCI account using your GitHub username and the basic auth password you just created:

$ tensorci login

Create a New Project

To register your git repo as a TensorCI project, navigate to your project’s directory, and run:

$ tensorci init

This will create a .tensorci.yml config file in the root of your project with the following contents:

#
# Python TensorCI configuration file
#
model:         path/to/model/file
prepro_data:   module1.module2:function
train:         module1.module2:function
test:          module1.module2:function
predict:       module1.module2:function
reload_model:  module1.module2:function

These config values will need to be modified to fit your project, but not all of them need to be set in order to simply train your first model. The table below describes these config values in more depth, gives examples for each, and explains when each value is required (if at all).

Config Key Descriptions
Key Value Example Required For
model
Relative path to save model to and read
model from
 data/model/ Always
prepro_data
Path to module function used to
preprocess raw dataset before training
 src.dataset:prepro Training
train
Path to module function that trains
your model
 src.model:train Training
test
Path to module function that tests
your model’s accuracy after training
 src.model:test Optional
predict
Path to module function used to make
model predictions from behind an API
 src.model:predict Predictions
reload_model
Path to module function used to reload
latest model into memory when swapping
out old model
 src.model:reload Predictions

Once you’ve modified this config file to integrate with your project, go ahead and push these changes up to GitHub.

Note: Now that you’ve created your TensorCI project, you can easily navigate to its web dashboard counterpart at any time by running tensorci dash from the root of your project.

Congrats! That’s all it takes to set up a TensorCI project. The last thing you need to do before you’re ready to start training is create a TensorCI Dataset. Once that’s done, you’ll be ready to train your model on the TensorCI training cluster.