whole lake
response = openai.Completion.create(
engine="text-davinci-002",
prompt=prompt_text,
temperature=1,
max_tokens=300,
top_p=1,
frequency_penalty=0.7,
presence_penalty=0.1,
stop=["\n"],
)
won't return anything. and yes I have something as my prompt. is there some error I am missing? I don't get a error printed (python btw)
abstract mango
```py
response = openai.Completion.create(
engine="text-davinci-002",
...
Put the keyword “await” before the openai call
The OpenAI API for JavaScript loads the response asynchronously, so you have to use concurrency to get the result.
whole lake
Put the keyword “await” before the openai call
I will try that when I'm home.
Was this a recent change in the API?
abstract mango
I will try that when I'm home.
Was this a recent change in the API?
It’s always been like this.
Concurrency is the common method of implementing REST endpoints
whole lake
It’s always been like this.
no. It worked previously without awaiting
I've changed it to the following and yet it won't work ```py
async def ask(question, chat_log=None):
print("ask")
prompt_text = f'{chat_log}{restart_sequence}: {question}{start_sequence}'
print("about to think")
response = await openai.Completion.create(
engine="text-davinci-003",
prompt=prompt_text,
temperature=1,
max_tokens=300,
top_p=1,
frequency_penalty=0.7,
presence_penalty=0.1,
stop=["\n"],
)
print("about to return")
story = response['choices'][0]['text']
print("return")
return str(story)
Ah alright, just didn’t get the context before with the call being in an async function.
whole lake
Ah alright, just didn’t get the context before with the call being in an async f...
it wasn't in a async function before
it was a normal function
but since you can't await a normal function, I had to turn it into a async function
abstract mango
Oh maybe you’re using an engine in that?
OpenAI recently deprecated completely the use of engines
abstract mango
but how should I tell openai then what model to use?
const response = await openai.createCompletion({
model: "text-davinci-003",
prompt: "Say this is a test",
max_tokens: 7,
temperature: 0,
});
Use the model parameter
Instead of engine
Also make sure your node packages are updated
whole lake
should be. How can I double-check them?
whole lake
const response = await openai.createCompletion({
model: "text-davinci-003",
...
So just to make this clear. This would be ok?
async def ask(question, chat_log=None):
print("ask")
prompt_text = f'{chat_log}{restart_sequence}: {question}{start_sequence}'
print("about to think")
response = await openai.Completion.create(
model="text-davinci-003",
prompt=prompt_text,
temperature=1,
max_tokens=300,
top_p=1,
frequency_penalty=0.7,
presence_penalty=0.1,
stop=["\n"],
)
print("about to return")
story = response['choices'][0]['text']
print("return")
return str(story)
Run that command
abstract mango
So just to make this clear. This would be ok?
```py
async def ask(question, chat...
Looks good
abstract mango
Yeah update your packages and make sure they’re up to date
I assume you’re instantiating the openai object properly with a proper config object and such.
whole lake
npm-check-updates
so I just installed npm and that still isn't a valid command
abstract mango
so I just installed npm and that still isn't a valid command
Hmm, I’ll be right back, give me a few minutes.
whole lake
@abstract mango u there?
abstract mango
Oh yeah sorry, give me another few minutes here
abstract mango
<@456226577798135808> u there?
Oh also what do you have restart_sequence and start_sequence set as by the way?
whole lake
Oh also what do you have `restart_sequence` and `start_sequence` set as by the w...
these are okay
abstract mango
these are okay
Yeah just that what do you have them set as?
Seeing that they are variables set outside of the main function
whole lake
don't worry about them.
abstract mango
Fair enough
whole lake
they are set okay.
abstract mango
It's all good
whole lake
they are defined by my bots name and the user input. which again are stored in my DB so the AI knows what she should say and what the user said
sorry my english isn't on point today
abstract mango
No worries
whole lake
well I'm still clueless what could be causing my problems
abstract mango
well I'm still clueless what could be causing my problems
Okay maybe it's the stop parameter
Remove that.
whole lake
shouldn't be. I never changed that since I build that AI
you know that stop prevents the bot from generating new lines with tokes aka. blasting the heck out of my wallet?
abstract mango
Though it seems like it won't generate any text with the stop parameter
Oh also remove the await before the openai call.
whole lake
here is a quick example of how stuff is stored in my DB (it's in german so if you want I can quickly translate that): ```py
"\nUser: Wie geht's dir Rachel": "\n\nRachel: Mir geht's gut. Und dir?\n",
ohhhhhhhh
abstract mango
Yeah
whole lake
Oh also remove the await before the openai call.
so no async function?
abstract mango
Then keep the async for the def, remove the await for the openai call
Just how the code is implemented mainly.
whole lake
alright
whole lake
here is a quick example of how stuff is stored in my DB (it's in german so if yo...
^
AH
I got it working now
abstract mango
I got it working now
Nice nice
Yeah for some reason it's returning nothing on my end, but glad you got your code working
whole lake
that's the code now: ```py
async def ask(question, chat_log=None):
prompt_text = f'{chat_log}{restart_sequence} {question}{start_sequence}'
response = openai.Completion.create(
model="text-davinci-003",
prompt=prompt_text,
temperature=1,
max_tokens=300,
top_p=1,
frequency_penalty=0.7,
presence_penalty=0.1,
stop=["\n"],
)
story = response['choices'][0]['text']
return str(story)
and the call of the function: ```py
thought = await message.channel.send(await ask(question=msg_content, chat_log=chat_log),
reference=message)
yeah I had the same issue. It's such a pain when you don't get any error printed (at least that was my problem)
abstract mango
{
"choices": [
{
"finish_reason": "stop",
"index": 0,
"logprobs": null,
"text": ""
}
],
"created": 1669836117,
"id": "cmpl-6IMwXlZvF8ulKraj5cgBtzoJcpEp0",
"model": "text-davinci-003",
"object": "text_completion",
"usage": {
"prompt_tokens": 26,
"total_tokens": 26
}
}
This is the JSON response I get
So it's stopping right as when it starts responding
whole lake
that's weird
abstract mango
that's weird
I'd honestly say best route to take is to create a fine tuned model with some data you make that's formatted properly.
That way the model has a better chance of responding with what you need.
whole lake
well yes and no
I want my bot to learn by herself. hence why I have a DB hooked to her where she constantly saves her answers to improve herself. Kinda like machine learning.
I only gave her a structure of what she is
The rest is up to her to generate
abstract mango
Oh yeah, OpenAI does have the whole fine tune feature set, which is great from my own experience.
whole lake
agreed
abstract mango