Support for mbart models?/ Could we get the output logits of decoders before the beam search?

[leoozy]

The mbart model is implemented by hugging face

[byshiue]

It requires to modify the source codes.

[leoozy]

@byshiue Thank you for your reply. If I just use the python interface, could I get the logits output of the decoder?

[byshiue]

If you use decoder op, then the output is the results of transformer block. If you use decoding op, then you need to modify the op and FT source codes.

[leoozy]

@byshiue Thank you. I will try to implement the mbart model

[xiuzhilu]

@leoozy hi,I also want to use mbart model is implemented by hugging face, can you implement it for mbart model? and if yes, can you share it with me? Thank you very much.