[dynamo exporter] Fix dynamic shapes with DynamicCache

[titaiwangms]

Describe your changes

transformers introduces "Cache" objects, which is a custom class that contains kv_cache. However, in the current exporter, this becomes an exception of the pre-condition: input should have the same tree spec as the dynamic_shapes. For example, although DynamicCache is now traceable in exporter, it does not make sense to dynamic_shapes. Therefore, we can't use input tree spec to unflatten dynamic_shapes anymore. This PR deletes that, and simply iterates the input tree spec and turn any tuple to list to overcome the format mismatch caused by config.json.

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 in a follow-up PR.

(Optional) Issue link

[justinchuby]

Feel free to merge this before the IR change

[titaiwangms]

@jambayk I think CI is being weird. Would you help me merge it?

[xiaoyu-work]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 2 pipeline(s).

[xiaoyu-work]

onnxscript fusion test failed: https://github.com/microsoft/Olive/blob/main/test/unit_test/passes/onnx/test_onnxscript_fusion.py#L45. How will this test change onnxscript fusion behavior?

[titaiwangms]

onnxscript fusion test failed: https://github.com/microsoft/Olive/blob/main/test/unit_test/passes/onnx/test_onnxscript_fusion.py#L45. How will this test change onnxscript fusion behavior?

Thanks! It looks like TreeSpec is still wrong. I will need another fix on this.