英文:
Different answers from chatgpt-api and web interface
问题
I'm trying to integrate openai (==0.27.6) into my system, and it works kinda fine, however, the replies I'm getting through the API are totally worse than the answers from the web interface (https://chat.openai.com/chat) .
The way I'm using it with the API (python) is:
completions = openai.ChatCompletion.create(
model = 'gpt-3.5-turbo',
messages=[
{
'role': 'user',
'content': prompt,
}
]
)
At this version I'm not using other parameters, like temperature, however, previously with version 0.26.4 I used this:
completions = openai.Completion.create(
engine='text-davinci-002',
prompt=prompt,
max_tokens=4000,
n=3,
stop=None,
temperature=0.8
)
.
Do You guys have any idea how can I set the first example code to give similar answers to the web interface? There're many parameters that can be set, however, I did not find any documentation about the used values for the web interface.
Thanks.
英文:
I'm trying to integrate openai (==0.27.6) into my system, and it works kinda fine, however, the replies I'm getting through the API are totally worse than the answers from the web interface (https://chat.openai.com/chat) .
The way I'm using it with the API (python) is:
completitions = openai.ChatCompletion.create(
model = 'gpt-3.5-turbo',
messages=[
{
'role': 'user',
'content': prompt,
}
]
)
At this version I'm not using other parameters, like temperature, however, previously with version 0.26.4 I used this:
completitions = openai.Completion.create(\
engine='text-davinci-002',
prompt=prompt,
max_tokens=4000,
n=3,
stop=None,
temperature=0.8
)
.
Do You guys have any idea how can I set the first example code to give similar answers to the web interface? There' re many parameters that can be set, however, I did not find any documentation about the used values for the web interface.
Thanks.
答案1
得分: 1
你正在使用引擎 'text-davinci-002'，这是 GPT-3 的变体，而网络门户使用的是 GPT-3.5-turbo。
英文:
you're using engine='text-davinci-002' which is a variation of GPT-3, and the web portal uses GPT-3.5-turbo
通过集体智慧和协作来改善编程学习和解决问题的方式。致力于成为全球开发者共同参与的知识库，让每个人都能够通过互相帮助和分享经验来进步。
评论