Temperature param being ignored?

[rishsriv]

Hi there! Thank you for maintaining this package. It has made our lives so much simpler.

I wanted to check if the temperature parameter is somehow being ignored in the library?

We have an implementation that is something like this

resp, err := Openai.CreateChatCompletion(
	c,
	openai.ChatCompletionRequest{
		Model:       "gpt-3.5-turbo",
		Messages:    messages,
		MaxTokens:   600,
		Temperature: 0.0,
		Stop:        []string{"```"},
	},
)

But this is ignoring the temperature parameter and returning a different response every time. This issue does not seem to be present when using the Python library or making a CURL request to the OpenAI API.

[rishsriv]

To replicate on Replit or one's own machine

package main

import (
	"context"
	"fmt"
	openai "github.com/sashabaranov/go-openai"
)

func main() {
	client := openai.NewClient("YOUR OPENAI TOKEN")
	resp, err := client.CreateChatCompletion(
		context.Background(),
		openai.ChatCompletionRequest{
			Model: openai.GPT3Dot5Turbo,
			Messages: []openai.ChatCompletionMessage{
				{
					Role:    openai.ChatMessageRoleUser,
					Content: "The president was",
				},
			},
      MaxTokens: 128,
      Temperature: 0.0,
		},
	)

	if err != nil {
		fmt.Printf("ChatCompletion error: %v\n", err)
		return
	}

	fmt.Println(resp.Choices[0].Message.Content)
}

This gives a different answer every time, while using Python or curl results in the same answer

[vvatanabe]

@rishsriv @sashabaranov I investigated the cause of the problem. Please see here.
#334 (comment)

[rishsriv]

Thank you so much for investigating this, @vvatanabe !

[vvatanabe]

Duplicate #9

[vvatanabe]

This is a duplicate issue with #9, so please discuss it in #9 from now on.