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

num_return_sequences and num_beams > 1

#17
by tlphams - opened

Hi,
Does SantaCoder support returning multiple sequences? I have tried setting num_return_sequences = 4 and num_beams = 4, but the results is so different and not as good as num_return_sequences = 1 .

Hello @tlphams , could you please provide more context as to how you're using the model for generation? Including code and outputs would be very helpful!

Sign up or log in to comment