does it limit the length of outputs?

[nnakamura3]

When decoding, I noticed that the output caption length is less than 16 tokens. Can I change this maximum caption length?

[simaoh]

@nnakamura3 ,
The model sample functions use seq_length given by opt.seq_length, which as you can see in the training script , is set during training to be the maximum length of a sequence in the data found by the dataloader
https://github.com/yahoo/object_relation_transformer/blob/master/dataloader.py#L60-L63

This makes sense during training, since no larger sentence will be found in the data.

If you need to change the maximum length of a generated sequence on inference, for your application, you can do so byadding a seq_length parameter to your opts.py file.

Does this answer your question?