Add sympy Serialization #850
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
unit tests to ensure round trip of each datatype using both numeric and sympy inputs.
Co-authored-by: Tanuj Khattar <tanujkhattar@google.com>
Sympy expressions can now be serialized through a recursive proto.
The qpe_hubbard_model is too large to test and causes developer friction. We have disabled the tests for now
These tests are very large and will cause problems when running. These tests are now disabled.
Move the changes from bloq_finder.py to bloq_report_card.py. This way, only the serialization portion of the test is affected.
Rather than building on top of quantumlib#849, we keep the changes separate.
tanujkhattar
requested changes
Apr 15, 2024
qualtran/serialization/bloq.py
Outdated
Comment on lines
99
to
112
| def decompose_sympy(expr: sympy.Expr): | ||
| function = _get_sympy_function_type(expr) | ||
| operands = [] | ||
| if function == sympy_pb2.Function.NONE: | ||
| parameter = _get_sympy_operand(expr) | ||
| operands.append(sympy_pb2.Operand(parameter=parameter)) | ||
|
|
||
| else: | ||
| for term in expr.args: | ||
| inner_term = decompose_sympy(term) | ||
|
|
||
| operands.append(sympy_pb2.Operand(term=inner_term)) | ||
|
|
||
| return sympy_pb2.Term(function=function, operands=operands) |
There was a problem hiding this comment.
Please add all the sympy related method to a new file qualtran/serialization/sympy.py and add corresponding tests to qualtran/serialization/sympy_test.py
Also, please add roundtrip unit tests to convert different types of sympy expression to / from proto. There are virtually no tests for sympy serialization verification right now.
qualtran/serialization/bloq.py
Outdated
| return sympy_pb2.Parameter(const_irrat=expr.name) | ||
|
|
||
|
|
||
| def decompose_sympy(expr: sympy.Expr): |
There was a problem hiding this comment.
Can we call this method sympy_expr_to_proto ? Also add return type annotations
Suggested change
| def decompose_sympy(expr: sympy.Expr): | |
| def sympy_expr_to_proto(expr: sympy.Expr) -> sympy_pb2.Term: |
qualtran/serialization/bloq.py
Outdated
| return function(parameters) | ||
|
|
||
|
|
||
| def compose_sympy(expr: sympy.Expr): |
There was a problem hiding this comment.
I think the type annotations are incorrect and the expr should be a sympy_pb2.Term ? Please also rename the method to sympy_expr_from_proto to keep the naming consistent with other serialization methods. Also add return type annotations
tanujkhattar
requested changes
Apr 16, 2024
qualtran/bloqs/factoring/mod_mul.py
Outdated
Comment on lines
73
to
74
| def namespace(self) -> str: | ||
| return "qualtran." + self.__module__ |
There was a problem hiding this comment.
I don't think this should be needed now?
In [5]: from qualtran.bloqs.factoring.mod_mul import CtrlModMul
In [6]: CtrlModMul(5, 10, 10).__module__
Out[6]: 'qualtran.bloqs.factoring.mod_mul'
qualtran/serialization/sympy.py
Outdated
Comment on lines
221
to
222
| else: | ||
| raise NotImplementedError(f"{term.function} has not been fully implimented.") |
There was a problem hiding this comment.
Suggested change
| else: | |
| raise NotImplementedError(f"{term.function} has not been fully implimented.") | |
| raise NotImplementedError(f"{term.function} has not been fully implemented.") |
qualtran/serialization/sympy.py
Outdated
| return deserialized_parameter | ||
|
|
||
|
|
||
| def sympy_expr_from_proto(term: sympy_pb2.Term) -> Any: |
There was a problem hiding this comment.
Can we use a more concrete return type instead of Any ?
qualtran/serialization/sympy.py
Outdated
|
|
||
|
|
||
| def sympy_expr_from_proto(term: sympy_pb2.Term) -> Any: | ||
| """Deserialize a sympy expression.""" |
There was a problem hiding this comment.
Please add the caveats to the docstring. When does it raise an error? Which terms can it return?
qualtran/protos/sympy.proto
Outdated
| } | ||
|
|
||
| // Represents a constant, rational number. | ||
| message Fraction { |
There was a problem hiding this comment.
Let's rename this to Rational for consistency with sympy
qualtran/serialization/sympy.py
Outdated
| return sympy_pb2.Parameter(const_symbol=sympy_pb2.ConstSymbol.Infinity) | ||
| if isinstance(expr, sympy.core.numbers.ImaginaryUnit): | ||
| return sympy_pb2.Parameter(const_symbol=sympy_pb2.ConstSymbol.ImaginaryUnit) | ||
| else: |
There was a problem hiding this comment.
Else is not needed.
Suggested change
| else: |
qualtran/serialization/sympy.py
Outdated
Comment on lines
186
to
202
| if parameter_type == "symbol": | ||
| deserialized_parameter = sympy.symbols(serialized_parameter.symbol) | ||
| elif parameter_type == "const_int": | ||
| deserialized_parameter = serialized_parameter.const_int | ||
| elif parameter_type == "const_rat": | ||
| fraction = serialized_parameter.const_rat | ||
| numerator = _get_parameter(fraction.numerator) | ||
| denominator = _get_parameter(fraction.denominator) | ||
| deserialized_parameter = sympy.Rational(numerator, denominator) | ||
| elif parameter_type == "const_float": | ||
| deserialized_parameter = serialized_parameter.const_float | ||
| elif parameter_type == "const_symbol": | ||
| deserialized_parameter = _get_sympy_const_from_enum(serialized_parameter.const_symbol) | ||
| else: | ||
| raise TypeError(f"Type is not supported for {serialized_input}") | ||
|
|
||
| return deserialized_parameter |
There was a problem hiding this comment.
Directly return the deserialized_parameter and raise a TypeError if none of the if clauses are satisfied.
Suggested change
| if parameter_type == "symbol": | |
| deserialized_parameter = sympy.symbols(serialized_parameter.symbol) | |
| elif parameter_type == "const_int": | |
| deserialized_parameter = serialized_parameter.const_int | |
| elif parameter_type == "const_rat": | |
| fraction = serialized_parameter.const_rat | |
| numerator = _get_parameter(fraction.numerator) | |
| denominator = _get_parameter(fraction.denominator) | |
| deserialized_parameter = sympy.Rational(numerator, denominator) | |
| elif parameter_type == "const_float": | |
| deserialized_parameter = serialized_parameter.const_float | |
| elif parameter_type == "const_symbol": | |
| deserialized_parameter = _get_sympy_const_from_enum(serialized_parameter.const_symbol) | |
| else: | |
| raise TypeError(f"Type is not supported for {serialized_input}") | |
| return deserialized_parameter | |
| if parameter_type == "symbol": | |
| return sympy.symbols(serialized_parameter.symbol) | |
| if parameter_type == "const_int": | |
| return serialized_parameter.const_int | |
| if parameter_type == "const_rat": | |
| fraction = serialized_parameter.const_rat | |
| numerator = _get_parameter(fraction.numerator) | |
| denominator = _get_parameter(fraction.denominator) | |
| return sympy.Rational(numerator, denominator) | |
| if parameter_type == "const_float": | |
| return serialized_parameter.const_float | |
| if parameter_type == "const_symbol": | |
| return _get_sympy_const_from_enum(serialized_parameter.const_symbol) | |
| raise TypeError(f"Type is not supported for {serialized_input}") | |
| return deserialized_parameter |
qualtran/serialization/sympy_test.py
Outdated
Comment on lines
21
to
33
| @pytest.mark.parametrize( | ||
| 'expr', | ||
| [ | ||
| ( | ||
| sympy.parse_expr("5") | ||
| + sympy.symbols("x") | ||
| + sympy.parse_expr("1/2") | ||
| + sympy.pi | ||
| + sympy.parse_expr("2j") | ||
| ), | ||
| (sympy.parse_expr("(-b + sqrt(-4*a*c + b**2))/(2*a)")), | ||
| ], | ||
| ) |
There was a problem hiding this comment.
Let's not depend upon parse_expr for tests. Instead, let's create a list of expressions; potentially generated by small functions; which we'd like to test and parameterize over it. For example:
Suggested change
| @pytest.mark.parametrize( | |
| 'expr', | |
| [ | |
| ( | |
| sympy.parse_expr("5") | |
| + sympy.symbols("x") | |
| + sympy.parse_expr("1/2") | |
| + sympy.pi | |
| + sympy.parse_expr("2j") | |
| ), | |
| (sympy.parse_expr("(-b + sqrt(-4*a*c + b**2))/(2*a)")), | |
| ], | |
| ) | |
| x, N = sympy.Symbol('x', positive=True), sympy.Symbol('N') | |
| # These should return a `sympy_pb2.Parameter` proto object? | |
| sympy_parameters_to_test = [ | |
| # Only symbols | |
| sympy.Symbol('x'), | |
| sympy.Symbol('N'), | |
| sympy.Symbol('E'), | |
| # Sympy constants | |
| sympy.pi, sympy.Infinity, sympy.E, sympy.I # etc. | |
| # Integers, Floats, Rationals | |
| sympy.Integer(5), | |
| sympy.Float(0.1), | |
| sympy.Rational("1/2"), | |
| sympy.Rational('1/10'), | |
| ] | |
| sympy_expr_to_test = [ | |
| 5 * x + sympy.sqrt(N), | |
| # Add different types of functions that are supported, maybe in a single expression or multiple expressions. | |
| ] | |
| @pytest.mark.parametrize('expr', sympy_parameters_to_test + sympy_exprs_to_test) |
tanujkhattar
approved these changes
Apr 16, 2024
Call sympy serialization directly rather than calling it through bloq.
Addressed nit comments. Co-authored-by: Tanuj Khattar <tanujkhattar@google.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Adds Protos to hold sympy expressions an a nested structure. Currently the following basic operations are supported: multiplication, addition, power, and mod. More operations will be added in a follow-up cl.