Use of train=false in dl.py

[akshitac8]

Can you elaborate the use of parameter train=false passed to word2vec_embedding_layer( ) as default parameter. You have already used pre-trained embedding, so what happens when we use train=true?

[sachka]

Having set train to False means the value of each word vector (coming from a pretrained embedding matrix) will remain unchanged at every epoch, otherwise they change at each backpropagation. The train value is ignored (It is always True) if you learn the matrix of vectors from the scratch.

[GauravBh1010tt]

@sachka is right. When you set the train value as false, the number of trainable parameters are reduced. This could help to converge quickly.

[adityac8]

@akshitac8 @GauravBh1010tt If the description is satisfactory then can we close this one? 🎏