{'input_ids': tensor([[ 101, 7414, 7798, ..., 520, 136, 102],
[ 101, 3209, 2225, ..., 2357, 136, 102],
[ 101, 4868, 2786, ..., 3511, 136, 102],
...,
[ 101, 704, 686, ..., 4955, 8043, 102],
[ 101, 1914, 961, ..., 4158, 8043, 102],
[ 101, 7679, 1046, ..., 3836, 136, 102]], device='cuda:0'), 'token_type_ids': tensor([[0, 0, 0, ..., 1, 1, 1],
[0, 0, 0, ..., 1, 1, 1],
[0, 0, 0, ..., 1, 1, 1],
...,
[0, 0, 0, ..., 1, 1, 1],
[0, 0, 0, ..., 1, 1, 1],
[0, 0, 0, ..., 1, 1, 1]], device='cuda:0'), 'attention_mask': tensor([[1, 1, 1, ..., 1, 1, 1],
[1, 1, 1, ..., 1, 1, 1],
[1, 1, 1, ..., 1, 1, 1],
...,
[1, 1, 1, ..., 1, 1, 1],
[1, 1, 1, ..., 1, 1, 1],
[1, 1, 1, ..., 1, 1, 1]], device='cuda:0'), 'start_positions': tensor([129, 22, 109, 256, 116, 31, 206, 85, 43, 19, 198, 256, 88, 80,
256, 256, 256, 256, 200, 40, 256, 256, 256, 46, 147, 256, 153, 22,
256, 0, 46, 27], device='cuda:0'), 'end_positions': tensor([131, 28, 113, 256, 119, 35, 211, 91, 50, 21, 200, 256, 91, 84,
256, 256, 256, 256, 208, 44, 256, 256, 256, 51, 151, 256, 156, 43,
256, 5, 49, 29], device='cuda:0')}

QuestionAnsweringModelOutput(loss=tensor(5.5453, device='cuda:0', grad_fn=<DivBackward0>), start_logits=tensor([[-0.4229, -0.4230, -0.4232, ..., -0.4227, -0.4227, -0.4229],
[-0.4082, -0.4082, -0.4084, ..., -0.4081, -0.4080, -0.4082],
[-0.4131, -0.4133, -0.4131, ..., -0.4130, -0.4129, -0.4130],
...,
[-0.4040, -0.4041, -0.4042, ..., -0.4039, -0.4037, -0.4039],
[-0.4106, -0.4107, -0.4108, ..., -0.4105, -0.4104, -0.4105],
[-0.4109, -0.4111, -0.4114, ..., -0.4107, -0.4107, -0.4108]],
device='cuda:0', grad_fn=<SqueezeBackward1>), end_logits=tensor([[0.7144, 0.7139, 0.7141, ..., 0.7149, 0.7152, 0.7148],
[0.7109, 0.7107, 0.7109, ..., 0.7114, 0.7118, 0.7115],
[0.7363, 0.7362, 0.7359, ..., 0.7368, 0.7371, 0.7369],
...,
[0.7121, 0.7114, 0.7115, ..., 0.7132, 0.7130, 0.7127],
[0.7161, 0.7152, 0.7156, ..., 0.7165, 0.7170, 0.7166],
[0.7267, 0.7264, 0.7260, ..., 0.7271, 0.7276, 0.7272]],
device='cuda:0', grad_fn=<SqueezeBackward1>), hidden_states=None, attentions=None)
tensor(5.5453, device='cuda:0', grad_fn=<DivBackward0>)