Stack from ghstack:

• [ONNX] Fix param names (#50764) #50955 [ONNX] Fix param names ([ONNX] Fix param names #50764)
• [ONNX] Replace optional parameters of Resize with placeholder for ops13. (#50574) #50954 [ONNX] Replace optional parameters of Resize with placeholder for ops13. ([ONNX] Replace optional parameters of Resize with placeholder for ops13. #50574)
• [ONNX] Support gelu for fp16 export (#50487) #50911 [ONNX] Support gelu for fp16 export ([ONNX] Support gelu for fp16 export #50487)
• [ONNX] Enable _jit_pass_onnx_fold_if only when dynamic_axes is None (#50582) #50910 [ONNX] Enable _jit_pass_onnx_fold_if only when dynamic_axes is None ([ONNX] Enable _jit_pass_onnx_fold_if only when dynamic_axes is None #50582)
• [ONNX] Add logical_and, logical_or, logical_xor torch op support in pytorch exporter (#50570) #50909 [ONNX] Add logical_and, logical_or, logical_xor torch op support in pytorch exporter ([ONNX] Add logical_and, logical_or, logical_xor torch op support in pytorch exporter #50570)
• [ONNX] Add binary_cross_entropy_with_logits op to ONNX opset version 12 (#49675) #50908 [ONNX] Add binary_cross_entropy_with_logits op to ONNX opset version 12 ([ONNX] Add binary_cross_entropy_with_logits op to ONNX opset version 12 #49675)
• [ONNX] Update Reducesum operator for opset 13 (#50532) #50907 [ONNX] Update Reducesum operator for opset 13 ([ONNX] Update Reducesum operator for opset 13 #50532)
• [ONNX] Support opset13 Squeeze and Unsqueeze (#50150) #50906 [ONNX] Support opset13 Squeeze and Unsqueeze ([ONNX] Support opset13 Squeeze and Unsqueeze #50150)
• [ONNX] Use parameter values in onnx shape inference (#49706) #50905 [ONNX] Use parameter values in onnx shape inference ([ONNX] Use parameter values in onnx shape inference #49706)

Adds an additional run of onnx shape inference after constant folding, since initializer may have changed and affected shape inference.

Differential Revision: D26050881