Text Generation
Transformers
PyTorch
code
gpt2
custom_code
Eval Results
text-generation-inference
Inference Endpoints

Stopping criteria

#39
by Mariusbrm - opened

I'm looking for a way to stop the generation whenever my task/prompt is fullfilled. Except from stop_words and max_length is there another stopping criteria that could be used to stop the generation. More specifically I want to 'force' the model to generate an end_of_sequence token before the max_length is reached.

Any ideas ?

Sign up or log in to comment