-
Notifications
You must be signed in to change notification settings - Fork 2.4k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add Qiskit native QPY ParameterExpression serialization #13356
Conversation
With the release of symengine 0.13.0 we discovered a version dependence on the payload format used for serializing symengine expressions. This was worked around in Qiskit#13251 but this is not a sustainable solution and only works for symengine 0.11.0 and 0.13.0 (there was no 0.12.0). While there was always the option to use sympy to serialize the underlying symbolic expression (there is a `use_symengine` flag on `qpy.dumps` you can set to `False` to do this) the sympy serialzation has several tradeoffs most importantly is much higher runtime overhead. To solve the issue moving forward a qiskit native representation of the parameter expression object is necessary for serialization. This commit bumps the QPY format version to 13 and adds a new serialization format for ParameterExpression objects. This new format is a serialization of the API calls made to ParameterExpression that resulted in the creation of the underlying object. To facilitate this the ParameterExpression class is expanded to store an internal "replay" record of the API calls used to construct the ParameterExpression object. This internal list is what gets serialized by QPY and then on deserialization the "replay" is replayed to reconstruct the expression object. This is a different approach to the previous QPY representations of the ParameterExpression objects which instead represented the internal state stored in the ParameterExpression object with the symbolic expression from symengine (or a sympy copy of the expression). Doing this directly in Qiskit isn't viable though because symengine's internal expression tree is not exposed to Python directly. There isn't any method (private or public) to walk the expression tree to construct a serialization format based off of it. Converting symengine to a sympy expression and then using sympy's API to walk the expression tree is a possibility but that would tie us to sympy which would be problematic for Qiskit#13267 and Qiskit#13131, have significant runtime overhead, and it would be just easier to rely on sympy's native serialization tools. The tradeoff with this approach is that it does increase the memory overhead of the `ParameterExpression` class because for each element in the expression we have to store a record of it. Depending on the depth of the expression tree this also could be a lot larger than symengine's internal representation as we store the raw api calls made to create the ParameterExpression but symengine is likely simplifying it's internal representation as it builds it out. But I personally think this tradeoff is worthwhile as it ties the serialization format to the Qiskit objects instead of relying on a 3rd party library. This also gives us the flexibility of changing the internal symbolic expression library internally in the future if we decide to stop using symengine at any point. Fixes Qiskit#13252
One or more of the following people are relevant to this code:
|
Pull Request Test Coverage Report for Build 11708157805Warning: This coverage report may be inaccurate.This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.
Details
💛 - Coveralls |
The state of this is it's stuck on the overly complex expression example test I added in: 2b352d7 I'm thinking it might be worth approaching the deserialization in a slightly different approach instead of doing nested expressions recursively like it currently is. If anyone has bandwidth and can come up with a fix before me for that test, feel free to just push to the branch. I think when that test is passing we probably have enough coverage between that new test and the expression tests that already existed for circuits and pulse schedules (we might want to convert those expression tests to circuit equivalents as part of the pulse removal in 2.0). |
This commit changes how the deserialization works to use a postfix stack based approach. Operands are push on the stack and then popped off based on the operation being run. The result of the operation is then pushed on the stack. This handles nested objects much more cleanly than the recursion based approach because we just keep pushing on the stack instead of recursing, making the accounting much simpler. After the expression payload is finished being processed there will be a single value on the stack and that is returned as the final expression.
Ok, I fixed the issue from last week with: a903387 which switches to a stack based approach instead of a recursive one. I think this should be ready now. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks @mtreinish. After taking a look I think the PR looks good. I think we agree in the need for a Qiskit-native alternative for the symengine serialization even if there are some tradeoffs in terms of memory footprint. I added mostly documentation-related suggestions, some typos and some rewordings, and a few minor comments. The question I left in read_parameter_expression_v4
might be important, but I might have missed something there. Other than that, I was wondering if we should also communicate somehow that use_symengine
isn't useful anymore for QPY versions over 13?
Co-authored-by: Elena Peña Tapia <[email protected]>
This commit fixes a bug with handling the operand order of subtraction, division, and exponentiation. These operations are not commutative but the qpy deserialization code was treating them as such. So in cases where the argument order was reversed qpy was trying to flip the operands around for code simplicity and this would result in incorrect behavior. This commit fixes this by adding explicit op codes for the reversed sub, div, and pow and preserving the operand order correctly in these cases.
I added a sentence to the We can drop the argument in 2.0 potentially since 2.0 will have a minimum compatibility version of qpy format version 13 so nothing >=2.0 emits can support the |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I added a sentence to the
qpy.dump
function's docstring for theuse_symengine
argument to say it doesn't do anything in qpy 13. Were you thinking we needed an upgrade release note or something to documente it further?
I didn't have a fixed idea, but I think the docstring is a good place for the moment. I also wasn't pushing for a deprecation now, just wanted to raise the point to make sure we didn't forget to deal with it in the future.
Thanks a lot for applying the review comments so fast! I have taken a second look and cannot come up with any other potential issue, so on my side I think the PR is ready to be merged.
Summary
With the release of symengine 0.13.0 we discovered a version dependence on the payload format used for serializing symengine expressions. This was worked around in #13251 but this is not a sustainable solution and only works for symengine 0.11.0 and 0.13.0 (there was no 0.12.0). While there was always the option to use sympy to serialize the underlying symbolic expression (there is a
use_symengine
flag onqpy.dumps
you can set toFalse
to do this) the sympy serialzation has several tradeoffs most importantly is much higher runtime overhead. To solve the issue moving forward a qiskit native representation of the parameter expression object is necessary for serialization.This commit bumps the QPY format version to 13 and adds a new serialization format for ParameterExpression objects. This new format is a serialization of the API calls made to ParameterExpression that resulted in the creation of the underlying object. To facilitate this the ParameterExpression class is expanded to store an internal "replay" record of the API calls used to construct the ParameterExpression object. This internal list is what gets serialized by QPY and then on deserialization the "replay" is replayed to reconstruct the expression object. This is a different approach to the previous QPY representations of the ParameterExpression objects which instead represented the internal state stored in the ParameterExpression object with the symbolic expression from symengine (or a sympy copy of the expression). Doing this directly in Qiskit isn't viable though because symengine's internal expression tree is not exposed to Python directly. There isn't any method (private or public) to walk the expression tree to construct a serialization format based off of it. Converting symengine to a sympy expression and then using sympy's API to walk the expression tree is a possibility but that would tie us to sympy which would be problematic for #13267 and #13131, have significant runtime overhead, and it would be just easier to rely on sympy's native serialization tools.
The tradeoff with this approach is that it does increase the memory overhead of the
ParameterExpression
class because for each element in the expression we have to store a record of it. Depending on the depth of the expression tree this also could be a lot larger than symengine's internal representation as we store the raw api calls made to create the ParameterExpression but symengine is likely simplifying it's internal representation as it builds it out. But I personally think this tradeoff is worthwhile as it ties the serialization format to the Qiskit objects and APIs instead of relying on a 3rd party library. This also gives us the flexibility of changing the internal symbolic expression library internally in the future if we decide to stop using symengine at any point.Details and comments
Fixes #13252
TODO: