Limit the use of PreTrainedModel.device (#16935)

* Limit the use of PreTrainedModel.device

* Fix

src/transformers/generation_utils.py

@@ -502,7 +502,7 @@ class GenerationMixin:
         if is_input_ids and is_pad_token_in_inputs and is_pad_token_not_equal_to_eos_token_id:
             return inputs.ne(pad_token_id).long()
         else:
-            return torch.ones(inputs.shape[:2], dtype=torch.long, device=self.device)
+            return torch.ones(inputs.shape[:2], dtype=torch.long, device=inputs.device)
 
     def _prepare_encoder_decoder_kwargs_for_generation(
         self, inputs_tensor: torch.Tensor, model_kwargs, model_input_name: Optional[str] = None
@@ -532,13 +532,16 @@ class GenerationMixin:
         decoder_start_token_id: int = None,
         bos_token_id: int = None,
         model_kwargs: Optional[Dict[str, torch.Tensor]] = None,
+        device: torch.device = None,
     ) -> torch.LongTensor:
 
         if model_kwargs is not None and "decoder_input_ids" in model_kwargs:
             return model_kwargs.pop("decoder_input_ids")
         else:
             decoder_start_token_id = self._get_decoder_start_token_id(decoder_start_token_id, bos_token_id)
-            return torch.ones((batch_size, 1), dtype=torch.long, device=self.device) * decoder_start_token_id
+            if device is None:
+                device = self.device
+            return torch.ones((batch_size, 1), dtype=torch.long, device=device) * decoder_start_token_id
 
     def _get_decoder_start_token_id(self, decoder_start_token_id: int = None, bos_token_id: int = None) -> int:
         decoder_start_token_id = (
@@ -1177,6 +1180,7 @@ class GenerationMixin:
                 decoder_start_token_id=decoder_start_token_id,
                 bos_token_id=bos_token_id,
                 model_kwargs=model_kwargs,
+                device=inputs_tensor.device,
             )
         else:
             # if decoder-only then inputs_tensor has to be `input_ids`
@@ -1327,7 +1331,7 @@ class GenerationMixin:
             beam_scorer = BeamSearchScorer(
                 batch_size=batch_size,
                 num_beams=num_beams,
-                device=self.device,
+                device=inputs_tensor.device,
                 length_penalty=length_penalty,
                 do_early_stopping=early_stopping,
                 num_beam_hyps_to_keep=num_return_sequences,
@@ -1367,7 +1371,7 @@ class GenerationMixin:
             beam_scorer = BeamSearchScorer(
                 batch_size=batch_size * num_return_sequences,
                 num_beams=num_beams,
-                device=self.device,
+                device=inputs_tensor.device,
                 length_penalty=length_penalty,
                 do_early_stopping=early_stopping,
             )
@@ -1410,7 +1414,7 @@ class GenerationMixin:
                 batch_size=batch_size,
                 num_beams=num_beams,
                 max_length=stopping_criteria.max_length,
-                device=self.device,
+                device=inputs_tensor.device,
                 length_penalty=length_penalty,
                 do_early_stopping=early_stopping,
                 num_beam_hyps_to_keep=num_return_sequences,
@@ -1492,7 +1496,7 @@ class GenerationMixin:
                 constraints=final_constraints,
                 batch_size=batch_size,
                 num_beams=num_beams,
-                device=self.device,
+                device=inputs_tensor.device,
                 length_penalty=length_penalty,
                 do_early_stopping=early_stopping,
                 num_beam_hyps_to_keep=num_return_sequences,

src/transformers/modeling_utils.py

@@ -1157,7 +1157,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
 
         # Build new embeddings
         new_embeddings = nn.Embedding(new_num_tokens, old_embedding_dim)
-        new_embeddings.to(self.device, dtype=old_embeddings.weight.dtype)
+        new_embeddings.to(old_embeddings.weight.device, dtype=old_embeddings.weight.dtype)
 
         # initialize all new embeddings (in particular added tokens)
         self._init_weights(new_embeddings)
@@ -1228,7 +1228,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
         new_lm_head_shape = (old_lm_head_dim, new_num_tokens) if not transposed else (new_num_tokens, old_lm_head_dim)
         has_new_lm_head_bias = old_lm_head.bias is not None
         new_lm_head = nn.Linear(*new_lm_head_shape, bias=has_new_lm_head_bias)
-        new_lm_head = new_lm_head.to(self.device, dtype=old_lm_head.weight.dtype)
+        new_lm_head = new_lm_head.to(old_lm_head.weight.device, dtype=old_lm_head.weight.dtype)
 
         # initialize new lm head (in particular added tokens)
         self._init_weights(new_lm_head)