strange error when inferring on novel examples

[ysymyth]

I tried to run the pre-trained model interactively, with db_id=baseball_1.

When I run this question (from training set)
Find the full name and id of the college that has the most baseball players.
it runs normally.

But when I change the question to
return college that has the most baseball players

Strange thing happens: beam search returns CUDA assertion error as db_scope_update_idx.max() = 205 while m_field_masks.size(1) = 197. So the following scatter_add_ command cannot run.

TabularSemanticParsing/src/semantic_parser/decoding_algorithms.py

Line 218 in 0b094d3

m_field_masks.scatter_add_(index=db_scope_update_idx, src=db_scope_update_mask.long(), dim=1)

Any idea why this happens? Happy to provide more details for replicating this weird bug...

[eche043]

Hello I have the same error.
Have you found a solution please?