A Problem with pretrain function in the mian_transductive.py

The evaluation usually occurs at the end of the training, when the epoch number is equal to max_epoch. However, in the original file when (epoch + 1) % 200 == 0, the node_classification_evaluation function will be called. It doesn't take into account the case when max_epoch is not 200, which will result in a Runtime error. Replacing 200 with max_epoch could fix the problem.