programmatic_mode_lab

Motivation

To interact with GLADOS without having to access your browser, we are introducing our new Programatic mode. Within this mode, the user can declare their experiment by implementing a class which contains all of their code and configurations to be run by the system.

Objective

At the end of this lab, you will be able to utilize the given experiment template to run your experiment locally and on GLADOS, our Automating Science Pipeline

Setup

1. To activate the virtual environment, my_evn, run the following command:

python3 -m venv my_env
source my_env/bin/activate

[⭐] For windows users, run the following command instead:

python -m venv my_env
.\my_env\Scripts\activate 

2. To download all necessary requirements to run this lab, run the following command:

pip install -r requirements.txt

Tutorial

To learn how to start your own experiment, we will implement the function: y = mx + b.

1. Copy from the template

You are provided with the template file exp_template.py. Make a new file called my_slope_exp.py in the root directory. Copy the contents of exp_template.py into my_slope_exp.py.

2. Declare Hyperparameters

Hyperparameters are defined using a dictionary of string to tuple, where the tuple defines the start, stop, and step of the variable. For our implementation of y = mx + b, we will declare our hyperparameter as:

hyperparams = {
        #variable : (start, stop, step_size)
            "x" : (0, 50, 1),
            "b" : (2, 2, 0),
            "m" : (0.5, 0.5, 0)
        }

The snippet above declares the variables x, b, and m with the ranges:

• 0 <= x <= 50 with increments of 1
• 2 <= b <= 2 with increments of 0
• 0.5 <= m <= 0.5 with increments of 0

2.5 Declare csv name (Optional)

Say you want to save the output csv of the result under a different name other than "output.csv". You can do it by declaring the new filename in the call to the super class like so:

super().__init__(
                hyperparams,
                csv_name="My_new_name.csv")

3. Implement process_trial()

The process_trial() method takes in data, which is a dictionary and outputs a dictionary of results. To implement y = mx + b, we can simply do:

    def process_trial(self, data):
        y = (data["m"] * data["x"]) + data["b"]

        return {"y" : y}

The snippet above define the variable y to be equals to (m * x) + b, and returns y as a result entry. You can collect more results if needed for your experiment as long as process_trial() returns a dictionary of String to Any.

[🤔] Keep in mind that your inputs are collected automatically, you do not have to include m, x, or b in your function returns. You will have access to both inputs and outputs later for graphing

4. Implement graph_result()

The graph_result() function is given for you to implement graph with the data collected from the experiment. For this lab, you can use matplotlib to graph the results with x axis = "x" and y axis = "y = mx + b". This function will open a matplotlib window displaying the graph of the results.

[⭐] Keep in mind that the shape of data now looks something like this:

data = {
    'x' = [0, 1, 2, ..., 50],
    'm' = [0.5, 0.5, 0.5, ...],
    'b' = [2, 2, 2, ...],
    'y' = [2, 2.5, 3, 3.5, ...]
}

1. Include the matplotlib library at the top of your file.

import matplotlib.pyplot as plt

2. Set up the graph with matplotlib.

    plt.plot(
            data["x"],
            data["y"], 
            marker='o', 
            linestyle='--', 
            color='blue'
            )

    plt.xlabel("x")
    plt.ylabel("y")
    plt.title("test_graph")
    plt.grid(True) # Add a grid
        
    plt.show()

[🧐] Note that the function also provides you with the variable save_path, which you can use to save your graph. This is for when you are submiting your code files to GLADOS and would like the system to produce a graph alongside a csv detailing the results

5. Run your experiment

    python3 my_slope_exp.py

[⭐] For windows users, run the following command instead:

    python my_slope_exp.py