diff --git a/optimum/exporters/neuron/utils.py b/optimum/exporters/neuron/utils.py
index 2a8d3df22..0d9f863bf 100644
--- a/optimum/exporters/neuron/utils.py
+++ b/optimum/exporters/neuron/utils.py
@@ -30,8 +30,7 @@
     DIFFUSION_MODEL_VAE_DECODER_NAME,
     DIFFUSION_MODEL_VAE_ENCODER_NAME,
     ENCODER_NAME,
-    get_attention_scores_sd2,
-    get_attention_scores_sd15,
+    get_attention_scores_sd,
     get_attention_scores_sdxl,
 )
 from ...utils import (
@@ -54,10 +53,7 @@
             "Please update diffusers by running `pip install --upgrade diffusers`"
         )
     from diffusers import ControlNetModel, UNet2DConditionModel
-    from diffusers.models.attention_processor import (
-        Attention,
-        AttnProcessor,
-    )
+    from diffusers.models.attention_processor import Attention
 
 
 if TYPE_CHECKING:
@@ -388,7 +384,6 @@ def get_submodels_for_export_stable_diffusion(
         models_for_export.append((DIFFUSION_MODEL_TEXT_ENCODER_2_NAME, copy.deepcopy(text_encoder_2)))
 
     # U-NET
-    pipeline.unet.set_attn_processor(AttnProcessor())
     pipeline.unet.config.text_encoder_projection_dim = projection_dim
     # The U-NET time_ids inputs shapes depends on the value of `requires_aesthetics_score`
     # https://github.com/huggingface/diffusers/blob/v0.18.2/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl_img2img.py#L571
@@ -400,12 +395,9 @@ def get_submodels_for_export_stable_diffusion(
         if is_sdxl:
             logger.info("Applying optimized attention score computation for sdxl.")
             Attention.get_attention_scores = get_attention_scores_sdxl
-        elif "v1-5" in pipeline.config._name_or_path:
-            logger.info("Applying optimized attention score computation for stable diffusion 1.5.")
-            Attention.get_attention_scores = get_attention_scores_sd15
         else:
-            logger.info("Applying optimized attention score computation for stable diffusion 2.")
-            Attention.get_attention_scores = get_attention_scores_sd2
+            logger.info("Applying optimized attention score computation for stable diffusion.")
+            Attention.get_attention_scores = get_attention_scores_sd
     else:
         logger.warning(
             "You are not applying optimized attention score computation. If you want better performance, please"
diff --git a/optimum/neuron/utils/__init__.py b/optimum/neuron/utils/__init__.py
index e2253306d..2f305947f 100644
--- a/optimum/neuron/utils/__init__.py
+++ b/optimum/neuron/utils/__init__.py
@@ -52,8 +52,7 @@
     ],
     "model_utils": ["get_tied_parameters_dict", "tie_parameters"],
     "optimization_utils": [
-        "get_attention_scores_sd2",
-        "get_attention_scores_sd15",
+        "get_attention_scores_sd",
         "get_attention_scores_sdxl",
     ],
     "patching": [
@@ -105,8 +104,7 @@
     )
     from .model_utils import get_tied_parameters_dict, tie_parameters
     from .optimization_utils import (
-        get_attention_scores_sd2,
-        get_attention_scores_sd15,
+        get_attention_scores_sd,
         get_attention_scores_sdxl,
     )
     from .patching import (
diff --git a/optimum/neuron/utils/optimization_utils.py b/optimum/neuron/utils/optimization_utils.py
index 5c1c5e70c..feff8e98c 100644
--- a/optimum/neuron/utils/optimization_utils.py
+++ b/optimum/neuron/utils/optimization_utils.py
@@ -17,43 +17,8 @@
 import torch
 
 
-def get_attention_scores_sd15(self, query, key, attention_mask) -> torch.Tensor:
-    """Optimized attention for Stable Diffusion 1.5 UNET."""
-    dtype = query.dtype
-
-    if self.upcast_attention:
-        query = query.float()
-        key = key.float()
-
-    baddbmm_input = torch.empty(query.shape[0], query.shape[1], key.shape[1], dtype=query.dtype, device=query.device)
-    beta = 0
-
-    attention_scores = torch.baddbmm(
-        baddbmm_input,
-        query,
-        key.transpose(-1, -2),
-        beta=beta,
-        alpha=self.scale,
-    )
-    del baddbmm_input
-
-    # TODO: following line is supposed to give the same result and reduce unnecessary overhead(no attention mask)
-    # however the compiled model output is far off from the one on cpu, need to further investigate.
-    # attention_scores = self.scale * torch.bmm(query, key.transpose(-1, -2))  # -> bad perf, max diff: 5.696073055267334 (atol: 0.001)
-
-    if self.upcast_softmax:
-        attention_scores = attention_scores.float()
-
-    attention_probs = torch.nn.functional.softmax(attention_scores, dim=-1)
-    del attention_scores
-
-    attention_probs = attention_probs.to(dtype)
-
-    return attention_probs
-
-
-def get_attention_scores_sd2(self, query, key, attn_mask):
-    """Optimized attention for Stable Diffusion 2 UNET."""
+def get_attention_scores_sd(self, query, key, attn_mask):
+    """Optimized attention for Stable Diffusion UNET."""
     dtype = query.dtype
 
     if self.upcast_attention: