bi-att-flow
bi-att-flow copied to clipboard
Why does the model work when turn 400 to 740
When training , we limit the paragraph length to 400 to train. But when testing, we use 740 as the paragraph length to test. Is it too amazing that the model still work?