Does training the model is not necessary?

[naknak-choi]

Hi there!

I've been reviewing the code and noticed that I couldn't find the training code. Does this mean that when using the PCW method, training the model is not necessary? Is it enough to just prepend the context window to the attention mechanism?

[nirrai21]

Hi!
That's exactly right, one of the advantages of PCW is that no training is needed at all. Please feel free to read the paper for further information.

[naknak-choi]

Thank you!!