While this is true, and was a major advantage of LSTMs/GRUs, they still suffer from vanishing gradients. w.r.t proteins, our sequences often surpass 1500 amino

While this is true, and was a major advantage of LSTMs/GRUs, they still suffer from vanishing gradients. w.r.t proteins, our sequences often surpass 1500 amino acids and that is really tough for an LSTM to stably train on.