no metrics: BLEU or Rouge-L

[Adam Dziedzic]

For the text-to-table task, there is no code to compute the metrics. When I run your training, there is no possibility of assessing how well it went? What was your approach to compute the metrics?

[Xuechen Li]

The table2text example code should periodically generate samples, with which you could evaluate standard NLG metrics using this codebase.

This was documented here.

[Adam Dziedzic]

Okay, thanks.