Fix `ErnieMEmbeddings` device issue (#21726)

* remove .parameters()).device

* fix

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/models/ernie_m/modeling_ernie_m.py

@@ -77,7 +77,7 @@ class ErnieMEmbeddings(nn.Module):
             inputs_embeds = self.word_embeddings(input_ids)
         if position_ids is None:
             input_shape = inputs_embeds.size()[:-1]
-            ones = torch.ones(input_shape, dtype=torch.int64)
+            ones = torch.ones(input_shape, dtype=torch.int64, device=inputs_embeds.device)
             seq_length = torch.cumsum(ones, dim=1)
             position_ids = seq_length - ones
 
@@ -85,7 +85,6 @@ class ErnieMEmbeddings(nn.Module):
                 position_ids = position_ids + past_key_values_length
         # to mimic paddlenlp implementation
         position_ids += 2
-        position_ids = position_ids.to(next(self.position_embeddings.parameters()).device)
         position_embeddings = self.position_embeddings(position_ids)
         embeddings = inputs_embeds + position_embeddings
         embeddings = self.layer_norm(embeddings)