เริ่มต้นใช้งานเริ่มต้นใช้งานได้ฟรี

Change the probability scale

In this exercise, you will see the difference in the resulted sentence when using different values of temperature to scale the probability distribution.

The function generate_phrase() is an adaptation of the function you created before and is already loaded in the environment. It receives the parameters model with the pre-trained model, initial_text with the context text and temperature that is the value to scale the softmax() function.

แบบฝึกหัดนี้เป็นส่วนหนึ่งของหลักสูตร

Recurrent Neural Networks (RNNs) for Language Modeling with Keras

ดูคอร์ส

คำแนะนำการฝึกหัด

  • Store the list of temperatures to the temperatures variable.
  • Loop a variable temperature over the temperatures list.
  • Generate a phrase using the pre-loaded function generate_phrase().
  • Print the temperature and the generated sentence.

แบบฝึกหัดเชิงโต้ตอบแบบลงมือทำ

ลองทำแบบฝึกหัดนี้โดยเติมโค้ดตัวอย่างนี้ให้สมบูรณ์

# Define the initial text
initial_text = "Spock and me "

# Define a vector with temperature values
____ = [0.2, 0.8, 1.0, 3.0, 10.0]

# Loop over temperatures and generate phrases
for ____ in temperatures:
	# Generate a phrase
	phrase = ____(model, initial_text, temperature)
    
	# Print the phrase
	print('Temperature {0}: {1}'.format(____, ____))
แก้ไขและรันโค้ด