Set test Dream transformer to have 2 layers so that CPU offload can be properly tested

Author: dg845

tests/models/transformers/test_models_transformer_dream.py

@@ -60,7 +60,7 @@ def prepare_dummy_input(self, batch_size: int = 1, sequence_length: int = 48):
 
     def prepare_init_args_and_inputs_for_common(self):
         init_dict = {
-            "num_layers": 1,
+            "num_layers": 2,
             "attention_head_dim": 16,
             "num_attention_heads": 4,
             "num_attention_kv_heads": 2,