Codegenerierung und Erklärung

Prompt-Engineering mit der OpenAI-API

Fouad Trad

Machine Learning Engineer

Codegenerierung

Erstellen von Programmcode, um ein bestimmtes Problem zu lösen.
Unverzichtbar in jedem Bereich, der Softwarelösungen nutzt.
Das Verstehen von generiertem Code ist wichtig, um ihn effektiv nutzen zu können.

Symbol, das Codes auf verschiedenen Bildschirmen anzeigt.

Prompts zur Codegenerierung

Problembeschreibung
Programmiersprache
Format (Skript, Funktion, Klasse)

prompt = "Write a Python function that accepts a list of quarterly sales, and outputs the 
average sales per quarter."
print(get_response(prompt))

def calculate_average_sales(quarterly_sales):

    total_sales = sum(quarterly_sales)
    average_sales = total_sales / len(quarterly_sales)

    return average_sales

Input-Output-Beispiele

Beispiele für Modell-Inputs und -Outputs, um ein Programm zu erstellen, das diese abbildet.

examples = """
Input: [150000, 180000, 200000, 170000] -> Output: 175000.0
Input: [10000, 25000, 30000, 15000] -> Output: 20000.0
Input: [50000, 75000, 60000, 45000] -> Output: 57500.0
"""

prompt = f"""You are provided with input-output examples delimited by triple 
backticks for a python program that receives a list of quarterly sales data. 
Write code for this program.
```{examples}```"""
print(get_response(prompt))

Input-Output-Beispiele

Based on the provided input-output examples, it seems like you want to calculate the average of 
quarterly sales data for each set of input values. 


Here's a Python program that does that:

def calculate_average_quarterly_sales(sales_data):
    total_sales = sum(sales_data)
    num_quarters = len(sales_data)
    average_sales = total_sales / num_quarters
    return average_sales


print(calculate_average_quarterly_sales([150000, 180000, 200000, 170000]))  # Output: 175000.0
print(calculate_average_quarterly_sales([10000, 25000, 30000, 15000])) # Output: 20000.0
print(calculate_average_quarterly_sales([50000, 75000, 60000, 45000]))  # Output: 57500.0

Anpassung des Codes

Aufforderung des Modells, den Code den Anforderungen entsprechend anzupassen.

script = """
quarterly_sales = [150, 180, 200, 170]
total_sales = sum(quarterly_sales)
print("Total sales: ", total_sales)
"""

prompt = f"""Modify the script delimited by triple backticks to a function that we 
can call to compute the total sales given quarterly sales
```{script}```"""
print(get_response(prompt))

Anpassung des Codes

def compute_total_sales(quarterly_sales):
    total_sales = sum(quarterly_sales)
    return total_sales


quarterly_sales = [150, 180, 200, 170]
total_sales = compute_total_sales(quarterly_sales)
print("Total sales: ", total_sales)

Mehrere Code-Anpassungen

script = """
quarterly_sales = [150, 180, 200, 170]
total_sales = sum(quarterly_sales)
print("Total sales: ", total_sales)
"""

prompt = f"""Modify the script delimited by triple backticks as follows:
             - Let user input parameters interactively


           ```{script}```"""
print(get_response(prompt))

Mehrere Code-Anpassungen

script = """
quarterly_sales = [150, 180, 200, 170]
total_sales = sum(quarterly_sales)
print("Total sales: ", total_sales)
"""

prompt = f"""Modify the script delimited by triple backticks as follows:
             - Let user input parameters interactively
             - Make sure to verify inputs are positive, otherwise, display a 
             message for the user, and ask them to provide their input again.
           ```{script}```"""
print(get_response(prompt))

Mehrere Code-Anpassungen

quarterly_sales = []

for i in range(4):
    while True:
        sales = float(input("Enter quarterly sales for quarter {}: ".format(i+1)))
        if sales <= 0:
             print("Sales must be positive. Please try again.")
             continue
        quarterly_sales.append(sales)
        break

total_sales = sum(quarterly_sales)
print("Total sales: ", total_sales)

Code-Erklärung

Code kann schwer zu verstehen sein.
LLMs können verwendet werden, um Code zu erklären.

Bild mit komplizierten Codes auf einem Bildschirm.

Anforderungen die Code-Erklärung

code = """
def compute_average_sales_per_quarter(quarterly_sales):
    average_sales = sum(quarterly_sales) / len(quarterly_sales)
    return average_sales
"""

Angabe der Länge der Erklärung

prompt = f"""Explain in one sentence what the code delimited by triple backticks does
             ```{code}```"""
print(get_response(prompt))

The code calculates the average sales per quarter by summing up the values in the 
"quarterly_sales" list and dividing the sum by the number of elements in the list.

Detaillierte Erklärung des Codes

code = """
def compute_average_sales_per_quarter(quarterly_sales):
    average_sales = sum(quarterly_sales) / len(quarterly_sales)
    return average_sales
"""

Ausführliche Erklärung mit chain-of-thought Prompt

prompt = f"""Explain what the code delimited by triple backticks does. 
Let's think step by step.
```{code}```"""
print(get_response(prompt))

Detaillierte Erklärung des Codes

The code defines a function called `compute_average_sales_per_quarter` that takes in a list of `quarterly_sales` as a 
parameter.


Inside the function, it calculates the average sales by summing up all the values in the `quarterly_sales` list using 
the `sum()` function, and then dividing it by the length of the list using the `len()` function. 

The result is stored in the variable `average_sales`.


Finally, the function returns the calculated average sales value.


This code snippet can be used to calculate the average sales per quarter given a list of quarterly sales data.

Lass uns üben!

Prompt-Engineering mit der OpenAI-API