How to Set Temperature in Langchain for Language Models

In Langchain, you set the temperature by passing the temperature parameter when creating a language model instance like OpenAI. The temperature controls randomness: lower values make output more focused, higher values make it more creative.

Syntax

To set temperature in Langchain, provide the temperature argument when initializing the language model. For example, with the OpenAI class, use OpenAI(temperature=0.7).

The temperature value is a float between 0 and 1 (sometimes higher), where 0 means very deterministic output and values closer to 1 increase randomness.

python

from langchain.llms import OpenAI

llm = OpenAI(temperature=0.7)

Example

This example shows how to create an OpenAI language model with a temperature of 0.9 and generate a text completion. Higher temperature makes the output more creative and varied.

python

from langchain.llms import OpenAI

# Create the language model with temperature 0.9
llm = OpenAI(temperature=0.9)

# Generate a completion
response = llm("Write a short, creative poem about the sun.")
print(response)

Output

The sun dances high, a golden flame,
Warming earth with gentle claim.
Bright rays weave through sky so blue,
A daily gift, forever new.

Common Pitfalls

Common mistakes include:

• Not setting the temperature parameter, which defaults to 1 and may produce very random output.
• Using values outside the typical range (0 to 1), which can cause unexpected behavior.
• Confusing temperature with other parameters like top_p or max_tokens.

Always check your model's documentation for supported temperature ranges.

python

from langchain.llms import OpenAI

# Wrong: temperature as string (will cause error)
# llm = OpenAI(temperature='high')

# Right: temperature as float
llm = OpenAI(temperature=0.5)

Quick Reference

Parameter	Description	Typical Values
temperature	Controls randomness of output	0.0 (deterministic) to 1.0 (creative)
top_p	Alternative to temperature for nucleus sampling	0.0 to 1.0
max_tokens	Maximum tokens to generate	Integer, e.g., 100
model_name	Name of the language model	e.g., 'gpt-4', 'text-davinci-003'

Key Takeaways

Set temperature in Langchain by passing the temperature parameter when creating the language model instance.

Temperature controls how creative or focused the output is: lower means more focused, higher means more creative.

Use float values between 0 and 1 for temperature; avoid invalid types or out-of-range values.

Check your model's documentation for supported temperature ranges and defaults.

Temperature is different from other parameters like top_p or max_tokens.