I tried to implement it and i didn’t know if I need to pre process the text first before feeding into the model.
Hi @Hunsen_Drake, To run an MBART model in Android using Java, you must first train the MBART model in TensorFlow and then convert it to TensorFlow Lite using the TFLite Converter. You cannot feed input text directly to the model. So, Before feeding input text, you need to implement the same tokenization process used during training to convert the raw input text string into a tensor of token IDs. Since MBART is a seq2seq model, you need to handle the input and output shapes as defined by the model. Thank you!