Consider an LSTM network with one hidden layer of 20 nodes used for predicting the next word in a text corpus. No bias is used in any of the nodes. The corpus is of length 1000 words and there are 100 unique words. Assume a 10 dimensional word embedding module outside of the LSTM network, whose output is fed to the word predictor LSTM network.

1. How many output nodes will be needed in the LSTM network, if one-hot encoding is used to represent a word? How many input nodes?
2. What will be the total number of trainable weights in the LSTM network? (Show individual values for all LSTM components for partial marking)