🧹 Remove deprecated RotaryEmbedding parts in the Attention layers (#34858)

* update

* style

* fix missing args

* remove last trace of old rope classes

* remove deprecated copied from

* fix copies

* trigger CIs

* post rebase clean-up

* reverse mistral

* cleanup after dropping commits

* Add comment

tests/models/stablelm/test_modeling_stablelm.py

@@ -44,11 +44,7 @@ if is_torch_available():
         StableLmForTokenClassification,
         StableLmModel,
     )
-    from transformers.models.stablelm.modeling_stablelm import (
-        StableLmDynamicNTKScalingRotaryEmbedding,
-        StableLmLinearScalingRotaryEmbedding,
-        StableLmRotaryEmbedding,
-    )
+    from transformers.models.stablelm.modeling_stablelm import StableLmRotaryEmbedding
 
 
 # Copied from transformers.tests.models.persimmon.test_modeling_persimmon.PersimmonModelTester with Persimmon -> StableLm
@@ -436,11 +432,12 @@ class StableLmModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterM
 
         # Sanity check linear RoPE scaling
         # New position "x" should match original position with index "x/scaling_factor"
-        linear_scaling_rope = StableLmLinearScalingRotaryEmbedding(
+        linear_scaling_rope = StableLmRotaryEmbedding(
             head_dim,
             max_position_embeddings=config.max_position_embeddings,
             base=config.rope_theta,
             scaling_factor=scaling_factor,
+            rope_type="linear",
         ).to(torch_device)
         linear_cos_short, linear_sin_short = linear_scaling_rope(x, position_ids_short)
         linear_cos_long, linear_sin_long = linear_scaling_rope(x, position_ids_long)
@@ -454,11 +451,12 @@ class StableLmModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterM
         # Sanity check Dynamic NTK RoPE scaling
         # Scaling should only be observed after a long input is fed. We can observe that the frequencies increase
         # with scaling_factor (or that `inv_freq` decreases)
-        ntk_scaling_rope = StableLmDynamicNTKScalingRotaryEmbedding(
+        ntk_scaling_rope = StableLmRotaryEmbedding(
             head_dim,
             max_position_embeddings=config.max_position_embeddings,
             base=config.rope_theta,
             scaling_factor=scaling_factor,
+            rope_type="dynamic",
         ).to(torch_device)
         ntk_cos_short, ntk_sin_short = ntk_scaling_rope(x, position_ids_short)
         ntk_cos_long, ntk_sin_long = ntk_scaling_rope(x, position_ids_long)