def test_fill_mask(self):
tokenizer = BigBirdTokenizer.from_pretrained("google/bigbird-roberta-base")
model = BigBirdForMaskedLM.from_pretrained("google/bigbird-roberta-base")
model.to(torch_device)
input_ids = tokenizer("The goal of life is [MASK] .", return_tensors="pt").input_ids.to(torch_device)
logits = model(input_ids).logits
# [MASK] is token at 6th position
pred_token = tokenizer.decode(torch.argmax(logits[0, 6:7], axis=-1))
self.assertEqual(pred_token, "happiness")
def create_and_check_for_masked_lm(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = BigBirdForMaskedLM(config=config)
model.to(torch_device)
model.eval()
result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=token_labels)
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size))