gpt-4o-mini

[butalex11]

Any plans for implement gpt-4o-mini support ?

[messenjah]

I tried to just add it to the models list in openai_helper.py on line 31
GPT_4O_MODELS = ("gpt-4o", "gpt-4o-mini")
but it fails when it uses tiktoken.

I assume tiktoken has to integrate the new model first.

[prttp]

I tried to just add it to the models list in openai_helper.py on line 31 GPT_4O_MODELS = ("gpt-4o", "gpt-4o-mini","gpt-4o-mini-2024-07-18") but it fails when it uses tiktoken.

I assume tiktoken has to integrate the new model first.

I made a workaround in openai_helper.py:654

            if (model == "gpt-4o-mini"):
                encoding = tiktoken.encoding_for_model("gpt-4o")
            else:
                encoding = tiktoken.encoding_for_model(model)

It is functional, so I'm planning to use it until tiktoken update

[messenjah]

I tried to just add it to the models list in openai_helper.py on line 31 GPT_4O_MODELS = ("gpt-4o", "gpt-4o-mini","gpt-4o-mini-2024-07-18") but it fails when it uses tiktoken.
I assume tiktoken has to integrate the new model first.

I made a workaround in openai_helper.py:654

            if (model == "gpt-4o-mini"):
                encoding = tiktoken.encoding_for_model("gpt-4o")
            else:
                encoding = tiktoken.encoding_for_model(model)

It is functional, so I'm planning to use it until tiktoken update

Hi prttp,

thanks for this quick answer and a work around!
Could you please share the exact changes you did? I tried to change it but it's still not working for me. I get the same encoding error with tiktoken. Here is how my block looks for me starting at line 660:

        model = self.config['model']
        try:
            if (model == "gpt-4o-mini"):
                encoding = tiktoken.encoding_for_model("gpt-4o")
            else:
                encoding = tiktoken.encoding_for_model(model)
        except KeyError:
            encoding = tiktoken.get_encoding("gpt-3.5-turbo")

I assume I did something wrong?

[prttp]

I tried to just add it to the models list in openai_helper.py on line 31 GPT_4O_MODELS = ("gpt-4o", "gpt-4o-mini","gpt-4o-mini-2024-07-18") but it fails when it uses tiktoken.
I assume tiktoken has to integrate the new model first.

I made a workaround in openai_helper.py:654

            if (model == "gpt-4o-mini"):
                encoding = tiktoken.encoding_for_model("gpt-4o")
            else:
                encoding = tiktoken.encoding_for_model(model)

It is functional, so I'm planning to use it until tiktoken update

Hi prttp,

thanks for this quick answer and a work around! Could you please share the exact changes you did? I tried to change it but it's still not working for me. I get the same encoding error with tiktoken. Here is how my block looks for me starting at line 660:

        model = self.config['model']
        try:
            if (model == "gpt-4o-mini"):
                encoding = tiktoken.encoding_for_model("gpt-4o")
            else:
                encoding = tiktoken.encoding_for_model(model)
        except KeyError:
            encoding = tiktoken.get_encoding("gpt-3.5-turbo")

I assume I did something wrong?

It should work, u can check my fork, I did exactly the same. U mentioned line 660 - do u use the last version of the repo? Which version of tiktoken is written in your requirements.txt?

[messenjah]

$ git pull origin main
From https:/n3d1117/chatgpt-telegram-bot
 * branch            main       -> FETCH_HEAD
Already up to date.

requirements.txt:3
tiktoken==0.7.0

Still not working. I will try to setup a new instance from scratch and try again

[messenjah]

I just setup a new instance from scratch and did the same changes and now it works! (So it was something else on my side)
Many thanks for this!