Shape inference: ReduceMean dispatcher, quant_pre_process: skip_symbo…

…lic_shape bugfix (microsoft#23558)

### Description
- Add symbolic shape inference dispatcher for `ReduceMean`.
- Reducemean is used in RMSNorm so shape inference fails for llama, phi,
etc torch exported models.
- Reuse the dispatcher for ReduceSum since ReduceMean 18+ and ReduceSum
13+ have the same specs other than the type of reduction done.
- Fix an issue with `quant_pre_process` tool where the external data
file is missing if `skip_symbolic_shape=True` and
`skip_optimization=False`.
- Add `"session.optimized_model_external_initializers_file_name"` to
session options so that the external data gets saved in the same temp
directory as the optimized model.



### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->

onnxruntime/python/tools/quantization/shape_inference.py

@@ -119,6 +119,12 @@ def quant_pre_process(
                     external_names, external_values = extract_raw_data_from_model(input_model)
                     sess_option.add_external_initializers(list(external_names), list(external_values))
                     input_model = input_model.SerializeToString()
+                # the saved optimized model otherwise points to the original external data file name
+                # which is not available relative to the optimized model file
+                elif skip_symbolic_shape and save_as_external_data:
+                    sess_option.add_session_config_entry(
+                        "session.optimized_model_external_initializers_file_name", "optimized.onnx.data"
+                    )
 
                 sess = onnxruntime.InferenceSession(input_model, sess_option, providers=["CPUExecutionProvider"])
                 # Close the session to avoid the cleanup error on Windows for temp folders

onnxruntime/python/tools/symbolic_shape_infer.py

@@ -166,6 +166,7 @@ def __init__(self, int_max, auto_merge, guess_output_rank, verbose, prefix=""):
             "Range": self._infer_Range,
             "Reciprocal": self._pass_on_shape_and_type,
             "ReduceSum": self._infer_ReduceSum,
+            "ReduceMean": self._infer_ReduceMean,
             "ReduceProd": self._infer_ReduceProd,
             "Reshape": self._infer_Reshape,
             "Resize": self._infer_Resize,
@@ -1603,6 +1604,11 @@ def _infer_ReduceSum(self, node):  # noqa: N802
                     )
                 )
 
+    def _infer_ReduceMean(self, node):  # noqa: N802
+        if get_opset(self.out_mp_) >= 18:
+            # reduce mean spec 18+ is same as reduce sum spec 13+
+            self._infer_ReduceSum(node)
+
     def _infer_ReduceProd(self, node):  # noqa: N802
         axes = get_attribute(node, "axes")
         keep_dims = get_attribute(node, "keepdims", 1)