Adapter new type promotion rule for Paddle 2.6 (PaddlePaddle#8079)

paddlenlp/layers/crf.py

@@ -198,7 +198,7 @@ def _trans_score(self, labels, lengths):
         flattened_transition_indices = transition_indices.reshape([-1])
         flattened_transition_params = paddle.flatten(self.transitions)
         scores = paddle.gather(flattened_transition_params, flattened_transition_indices).reshape([batch_size, -1])
-        mask_scores = scores * mask[:, 1:]
+        mask_scores = scores * mask[:, 1:].astype(scores.dtype)
 
         # Accumulate the transition score
         score = paddle.sum(mask_scores, 1)

paddlenlp/transformers/bloom/modeling.py

@@ -153,7 +153,9 @@ def build_alibi_tensor(attention_mask: Tensor, num_heads: int, dtype) -> Tensor:
     # => the query_length dimension will then be broadcasted correctly
     # This is more or less identical to T5's relative position bias:
     # https://github.com/huggingface/transformers/blob/f681437203baa7671de3174b0fa583c349d9d5e1/src/transformers/models/t5/modeling_t5.py#L527
-    arange_tensor = ((attention_mask.astype(paddle.float32).cumsum(axis=-1) - 1) * attention_mask)[:, None, :]
+    arange_tensor = (
+        (attention_mask.astype(paddle.float32).cumsum(axis=-1) - 1) * attention_mask.astype(paddle.float32)
+    )[:, None, :]
     alibi = slopes[..., None] * arange_tensor
     # return alibi
     return paddle.cast(alibi, dtype)

paddlenlp/transformers/electra/modeling.py

@@ -1051,7 +1051,7 @@ def get_discriminator_inputs(self, inputs, raw_inputs, generator_logits, generat
         mask_positions = paddle.where(generator_labels == -100, umask_positions, mask_positions)
         updated_inputs = self.update_inputs(inputs, sampled_tokids, mask_positions)
         # use inputs and updated_input to get discriminator labels
-        labels = mask_positions * (paddle.ones_like(inputs) - paddle.equal(updated_inputs, raw_inputs).astype("int32"))
+        labels = mask_positions * (paddle.ones_like(inputs) - paddle.equal(updated_inputs, raw_inputs).astype("int64"))
         return updated_inputs, labels, sampled_tokids
 
     def sample_from_softmax(self, logits, use_softmax_sample=True):