diff --git a/pytorch_lightning/models/trainer.py b/pytorch_lightning/models/trainer.py
index 8ee004adb3..655316ed3a 100644
--- a/pytorch_lightning/models/trainer.py
+++ b/pytorch_lightning/models/trainer.py
@@ -363,9 +363,6 @@ class Trainer(TrainerIO):
         # filter out the weights that were done on gpu so we can load on good old cpus
         self.optimizers = model.configure_optimizers()
 
-        # attach model to DP
-        model = LightningDataParallel(model, device_ids=self.data_parallel_device_ids)
-
         # run through amp wrapper
         if self.use_amp:
             # An example
@@ -374,6 +371,9 @@ class Trainer(TrainerIO):
             )
             self.optimizers = optimizers
 
+        if self.on_gpu:
+            model = LightningDataParallel(model, device_ids=self.data_parallel_device_ids)
+
         self.__run_pretrain_routine(model)
 
     def ddp_train(self, gpu_nb, model):