problem collection I’ve met and solved，遇到的问题及解决。

[momozhangcn]

1. when preprocess.py
   P: TypeError: reduce_ex() takes exactly one argument (0 given)
   W: Replace python3.7 with 3.6
2. when translate.py
   P:RuntimeError: index_select(): Expected dtype int32 or int64 for index
   W: change ALL ”select_indices = batch_index.view(-1) ” into ”select_indices = batch_index.view(-1).int()”
3. when score_predictions.py
   P: ValueError("Length of values does not match length of index")
   W: the arguments(-beam_size and -n_best) in preprocess.py must ＞the arguments(-beam_size) in score_predictions.py,finally we can get top 1,3,5,10 accuracy as we want
   BTW: my environment is python3.6(≤3.6)，pytorch1.8(it doesn't matter)