Fix onnx 1.19.0 issue in graph surgery (microsoft#2129)

## Describe your changes

1. Fix onnx 1.19.0 issue in graph surgery
1. Pin onnx version in test requirements.txt

## Checklist before requesting a review
- [ ] Add unit tests for this change.
- [ ] Make sure all tests can pass.
- [ ] Update documents if necessary.
- [ ] Lint and apply fixes to your code by running `lintrunner -a`
- [ ] Is this a user-facing change? If yes, give a description of this
change to be included in the release notes.
- [ ] Is this PR including examples changes? If yes, please remember to
update [example
documentation](https://github.com/microsoft/Olive/blob/main/docs/source/examples.md)
in a follow-up PR.

## (Optional) Issue link

Author: xiaoyu-work

olive/passes/onnx/graph_surgeries.py

@@ -247,8 +247,7 @@ def __init__(self, node_name, input_idx):
         self.input_idx = input_idx
 
     def __call__(self, model: ModelProto):
-        from onnx.helper import make_node
-        from onnx.mapping import TENSOR_TYPE_TO_NP_TYPE
+        from onnx.helper import make_node, tensor_dtype_to_np_dtype
 
         node = self.get_node_by_name(model, self.node_name)
         if node is None:
@@ -292,7 +291,7 @@ def __call__(self, model: ModelProto):
         if target_shape is None or target_type is None:
             raise ValueError(f"Cannot determine shape and type for input '{input_name}'.")
 
-        zero_values = np.zeros(target_shape, dtype=TENSOR_TYPE_TO_NP_TYPE[target_type])
+        zero_values = np.zeros(target_shape, dtype=tensor_dtype_to_np_dtype(target_type))
 
         zero_tensor = make_tensor(
             name=f"{self.node_name}_zero_tensor",

test/requirements-test.txt

@@ -16,7 +16,7 @@ neural-compressor<2.4
 nncf>=2.16.0
 numpy<2.0.0
 nvidia-modelopt
-onnx>=1.18.0
+onnx==1.18.0
 onnx-graphsurgeon
 onnxconverter_common
 onnxmltools