aruggero

Hi @navneet1v , thank you for your answer. I understand what you are saying but I still have some doubts. **1. Index uses NMSLib** > Index uses NMSLib: > While...

Thanks @frankfliu I've tried with the huggingface model. The answer is better, even if I'm still getting separator tokens: `[SEP] bbc japan was a general entertainment channel. which operated between...

Another question.. I see that the implementation of `ai.djl.huggingface.translator.QuestionAnsweringTranslator` is a bit different from the one suggested here https://docs.djl.ai/master/docs/demos/jupyter/pytorch/load_your_own_pytorch_bert.html I've this custom translator implemented in order to use this [HuggingFace...

Yes, in python it returns the first part of the context as the answer. Therefore: "BBC Japan". This is fine since the question is ambiguous "BBC Japan broadcasting". It returns...

Hi @frankfliu sorry for the delay Is this analysis necessary? We have the problem also with the model given by the DJL.. the distilbert: ``` public static ZooModel loadPyTorchModel() throws...

Hi @frankfliu Do you use the code I put at the top of the issue? The question I used is a bit different from the original one. I used "BBC...

Hi @frankfliu From what I know it is important to pass an input that reflects the one the model has seen during training (therefore with special tokens). What does removing...