diff --git a/configs/olive/sd/text_encoder.json b/configs/olive/sd/text_encoder.json
index 3e4aeb32b..91a3d57d3 100644
--- a/configs/olive/sd/text_encoder.json
+++ b/configs/olive/sd/text_encoder.json
@@ -36,6 +36,17 @@
     }
   },
   "passes": {
+    "optimize_CPUExecutionProvider": {
+      "type": "OrtTransformersOptimization",
+      "disable_search": true,
+      "config": {
+        "model_type": "clip",
+        "opt_level": 0,
+        "float16": false,
+        "use_gpu": false,
+        "keep_io_types": false
+      }
+    },
     "optimize_DmlExecutionProvider": {
       "type": "OrtTransformersOptimization",
       "disable_search": true,
@@ -91,6 +102,17 @@
         "use_gpu": true,
         "keep_io_types": false
       }
+    },
+    "quantization": {
+      "type": "OnnxDynamicQuantization",
+      "disable_search": true,
+      "config": {
+        "save_as_external_data": false,
+        "all_tensors_to_one_file": true,
+        "per_channel": false,
+        "reduce_range": false,
+        "MatMulConstBOnly": true
+      }
     }
   },
   "pass_flows": [["optimize"]],
diff --git a/configs/olive/sd/unet.json b/configs/olive/sd/unet.json
index 1b02ce0c3..907504b06 100644
--- a/configs/olive/sd/unet.json
+++ b/configs/olive/sd/unet.json
@@ -53,6 +53,17 @@
     }
   },
   "passes": {
+    "optimize_CPUExecutionProvider": {
+      "type": "OrtTransformersOptimization",
+      "disable_search": true,
+      "config": {
+        "model_type": "clip",
+        "opt_level": 0,
+        "float16": false,
+        "use_gpu": false,
+        "keep_io_types": false
+      }
+    },
     "optimize_DmlExecutionProvider": {
       "type": "OrtTransformersOptimization",
       "disable_search": true,
@@ -108,6 +119,17 @@
         "use_gpu": true,
         "keep_io_types": false
       }
+    },
+    "quantization": {
+      "type": "OnnxDynamicQuantization",
+      "disable_search": true,
+      "config": {
+        "save_as_external_data": false,
+        "all_tensors_to_one_file": true,
+        "per_channel": false,
+        "reduce_range": false,
+        "MatMulConstBOnly": true
+      }
     }
   },
   "pass_flows": [["optimize"]],
diff --git a/configs/olive/sd/vae_decoder.json b/configs/olive/sd/vae_decoder.json
index 3653ad422..2e21c0f79 100644
--- a/configs/olive/sd/vae_decoder.json
+++ b/configs/olive/sd/vae_decoder.json
@@ -43,6 +43,17 @@
     }
   },
   "passes": {
+    "optimize_CPUExecutionProvider": {
+      "type": "OrtTransformersOptimization",
+      "disable_search": true,
+      "config": {
+        "model_type": "clip",
+        "opt_level": 0,
+        "float16": false,
+        "use_gpu": false,
+        "keep_io_types": false
+      }
+    },
     "optimize_DmlExecutionProvider": {
       "type": "OrtTransformersOptimization",
       "disable_search": true,
@@ -98,6 +109,17 @@
         "use_gpu": true,
         "keep_io_types": false
       }
+    },
+    "quantization": {
+      "type": "OnnxDynamicQuantization",
+      "disable_search": true,
+      "config": {
+        "save_as_external_data": false,
+        "all_tensors_to_one_file": true,
+        "per_channel": false,
+        "reduce_range": false,
+        "MatMulConstBOnly": true
+      }
     }
   },
   "pass_flows": [["optimize"]],
diff --git a/configs/olive/sd/vae_encoder.json b/configs/olive/sd/vae_encoder.json
index 52dc369f9..2f012b187 100644
--- a/configs/olive/sd/vae_encoder.json
+++ b/configs/olive/sd/vae_encoder.json
@@ -43,6 +43,17 @@
     }
   },
   "passes": {
+    "optimize_CPUExecutionProvider": {
+      "type": "OrtTransformersOptimization",
+      "disable_search": true,
+      "config": {
+        "model_type": "clip",
+        "opt_level": 0,
+        "float16": false,
+        "use_gpu": false,
+        "keep_io_types": false
+      }
+    },
     "optimize_DmlExecutionProvider": {
       "type": "OrtTransformersOptimization",
       "disable_search": true,
@@ -98,6 +109,17 @@
         "use_gpu": true,
         "keep_io_types": false
       }
+    },
+    "quantization": {
+      "type": "OnnxDynamicQuantization",
+      "disable_search": true,
+      "config": {
+        "save_as_external_data": false,
+        "all_tensors_to_one_file": true,
+        "per_channel": false,
+        "reduce_range": false,
+        "MatMulConstBOnly": true
+      }
     }
   },
   "pass_flows": [["optimize"]],
diff --git a/modules/onnx_ep.py b/modules/onnx_ep.py
index 148d89840..6204a375c 100644
--- a/modules/onnx_ep.py
+++ b/modules/onnx_ep.py
@@ -15,7 +15,7 @@ class ExecutionProvider(str, Enum):
 
 available_execution_providers: List[ExecutionProvider] = ort.get_available_providers()
 EP_TO_NAME = {
-    ExecutionProvider.CPU: "cpu", # is this a valid option?
+    ExecutionProvider.CPU: "gpu-cpu", # ???
     ExecutionProvider.DirectML: "gpu-dml",
     ExecutionProvider.CUDA: "gpu-cuda", # test required
     ExecutionProvider.ROCm: "gpu-rocm", # test required
diff --git a/modules/onnx_pipelines.py b/modules/onnx_pipelines.py
index ce48fea9d..4f2af1e1b 100644
--- a/modules/onnx_pipelines.py
+++ b/modules/onnx_pipelines.py
@@ -244,30 +244,31 @@ class OnnxRawPipeline(OnnxPipelineBase):
                         if flow[i] == "optimize":
                             flow[i] = pass_key
                 olive_config["input_model"]["config"]["model_path"] = os.path.abspath(os.path.join(in_dir, submodel, "model.onnx"))
-                olive_config["passes"][pass_key]["config"]["float16"] = shared.opts.olive_float16
                 olive_config["engine"]["execution_providers"] = [shared.opts.onnx_execution_provider]
-                if shared.opts.onnx_execution_provider == ExecutionProvider.CUDA or shared.opts.onnx_execution_provider == ExecutionProvider.ROCm:
-                    if version.parse(ort.__version__) < version.parse("1.17.0"):
-                        olive_config["passes"][pass_key]["config"]["optimization_options"] = {"enable_skip_group_norm": False}
-                    if shared.opts.olive_float16:
-                        olive_config["passes"][pass_key]["config"]["keep_io_types"] = False
+                if pass_key in olive_config["passes"]:
+                    olive_config["passes"][pass_key]["config"]["float16"] = shared.opts.olive_float16
+                    if shared.opts.onnx_execution_provider == ExecutionProvider.CUDA or shared.opts.onnx_execution_provider == ExecutionProvider.ROCm:
+                        if version.parse(ort.__version__) < version.parse("1.17.0"):
+                            olive_config["passes"][pass_key]["config"]["optimization_options"] = {"enable_skip_group_norm": False}
+                        if shared.opts.olive_float16:
+                            olive_config["passes"][pass_key]["config"]["keep_io_types"] = False
 
                 run(olive_config)
 
                 with open(os.path.join("footprints", f"{submodel}_{EP_TO_NAME[shared.opts.onnx_execution_provider]}_footprints.json"), "r") as footprint_file:
                     footprints = json.load(footprint_file)
-                optimizer_footprint = None
+                processor_final_pass_footprint = None
                 for _, footprint in footprints.items():
-                    if footprint["from_pass"] == "OrtTransformersOptimization":
-                        optimizer_footprint = footprint
+                    if footprint["from_pass"] == olive_config["passes"][olive_config["pass_flows"][-1][-1]]["type"]:
+                        processor_final_pass_footprint = footprint
 
-                assert optimizer_footprint, "Failed to optimize model"
+                assert processor_final_pass_footprint, "Failed to optimize model"
 
                 optimized_model_paths[submodel] = ONNXModel(
-                    **optimizer_footprint["model_config"]["config"]
+                    **processor_final_pass_footprint["model_config"]["config"]
                 ).model_path
 
-                log.info(f"Optimized {submodel}")
+                log.info(f"Processed {submodel}")
 
             for submodel in self.submodels:
                 src_path = optimized_model_paths[submodel]