torch.fx.experimental.migrate_gradual_types.constraint_transformation.transform_get_item#
- torch.fx.experimental.migrate_gradual_types.constraint_transformation.transform_get_item(constraint, counter)[source]#
generate an equality of the form: t = [a1, …, an] then generate constraints that check if the given index is valid given this particular tensor size. If the index is valid, generate a constraint to get the item Note that we already handled the Dyn input case in the previous step. :param constraint: GetItem which assumes we are getting an item from a tensor (not Dyn) :param counter: variable tracking
Returns: simplified constraints for GetItem
