Text Generation
Transformers
PyTorch
Safetensors
Japanese
gpt_neox
lm
nlp
text-generation-inference

Maximum context lenght

#3
by IRITakeda - opened

What is the upper limit of maximum context lenght for japanese-gpt-neox-3.6b-instruction-ppo?
In the case of chatgpt, the total number of tokens in the prompt and response must be within the limit, is the same for Rinna-3.6B?

Yes it is the same as the pretrained 3.6B model, and the maximum length can also be found in the config file.
https://huggingface.co/rinna/japanese-gpt-neox-3.6b-instruction-ppo/blob/2140541486bfb31269acd035edd51208da40185b/config.json#L12

Thank you.
The number of tokens is different between ChatGPT and Rinna-3.6B, even though they have the same input.
Is there any way to calculate the number of tokens for Rinna-3.6B input?

Yes, you can check the length of token_ids as in the README example.

tianyuz changed discussion status to closed

Sign up or log in to comment