Hi Swati,

According to your first question, you can set maxlen argument to pad_sequence(), and than the Embedding input length set the same value.
( ref: https://www.tensorflow.org/api_docs/python/tf/keras/utils/pad_sequences#args )

About the question-2, you need to statistic your all sentences length distributed, and choose the mode value of all sentences length distributed.
( ref: https://www.carted.com/blog/variable-length-sequences-in-tensorflow-part-1/ )