├── .github └── FUNDING.yml ├── .gitignore ├── .pre-commit-config.yaml ├── LICENSE ├── MANIFEST.in ├── README.md ├── setup.py ├── sympytorch ├── __init__.py ├── hide_floats_m.py ├── py.typed └── sympy_module.py └── tests ├── test_hide_floats.py └── test_sympymodule.py /.github/FUNDING.yml: -------------------------------------------------------------------------------- 1 | github: [patrick-kidger] 2 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | **/__pycache__ 2 | *.egg-info 3 | build/ 4 | dist/ 5 | 6 | -------------------------------------------------------------------------------- /.pre-commit-config.yaml: -------------------------------------------------------------------------------- 1 | repos: 2 | - repo: https://github.com/ambv/black 3 | rev: 22.3.0 4 | hooks: 5 | - id: black 6 | - repo: https://github.com/charliermarsh/ruff-pre-commit 7 | rev: 'v0.0.255' 8 | hooks: 9 | - id: ruff 10 | args: ["--fix"] 11 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Apache License 2 | Version 2.0, January 2004 3 | http://www.apache.org/licenses/ 4 | 5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 6 | 7 | 1. Definitions. 8 | 9 | "License" shall mean the terms and conditions for use, reproduction, 10 | and distribution as defined by Sections 1 through 9 of this document. 11 | 12 | "Licensor" shall mean the copyright owner or entity authorized by 13 | the copyright owner that is granting the License. 14 | 15 | "Legal Entity" shall mean the union of the acting entity and all 16 | other entities that control, are controlled by, or are under common 17 | control with that entity. For the purposes of this definition, 18 | "control" means (i) the power, direct or indirect, to cause the 19 | direction or management of such entity, whether by contract or 20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the 21 | outstanding shares, or (iii) beneficial ownership of such entity. 22 | 23 | "You" (or "Your") shall mean an individual or Legal Entity 24 | exercising permissions granted by this License. 25 | 26 | "Source" form shall mean the preferred form for making modifications, 27 | including but not limited to software source code, documentation 28 | source, and configuration files. 29 | 30 | "Object" form shall mean any form resulting from mechanical 31 | transformation or translation of a Source form, including but 32 | not limited to compiled object code, generated documentation, 33 | and conversions to other media types. 34 | 35 | "Work" shall mean the work of authorship, whether in Source or 36 | Object form, made available under the License, as indicated by a 37 | copyright notice that is included in or attached to the work 38 | (an example is provided in the Appendix below). 39 | 40 | "Derivative Works" shall mean any work, whether in Source or Object 41 | form, that is based on (or derived from) the Work and for which the 42 | editorial revisions, annotations, elaborations, or other modifications 43 | represent, as a whole, an original work of authorship. For the purposes 44 | of this License, Derivative Works shall not include works that remain 45 | separable from, or merely link (or bind by name) to the interfaces of, 46 | the Work and Derivative Works thereof. 47 | 48 | "Contribution" shall mean any work of authorship, including 49 | the original version of the Work and any modifications or additions 50 | to that Work or Derivative Works thereof, that is intentionally 51 | submitted to Licensor for inclusion in the Work by the copyright owner 52 | or by an individual or Legal Entity authorized to submit on behalf of 53 | the copyright owner. For the purposes of this definition, "submitted" 54 | means any form of electronic, verbal, or written communication sent 55 | to the Licensor or its representatives, including but not limited to 56 | communication on electronic mailing lists, source code control systems, 57 | and issue tracking systems that are managed by, or on behalf of, the 58 | Licensor for the purpose of discussing and improving the Work, but 59 | excluding communication that is conspicuously marked or otherwise 60 | designated in writing by the copyright owner as "Not a Contribution." 61 | 62 | "Contributor" shall mean Licensor and any individual or Legal Entity 63 | on behalf of whom a Contribution has been received by Licensor and 64 | subsequently incorporated within the Work. 65 | 66 | 2. Grant of Copyright License. Subject to the terms and conditions of 67 | this License, each Contributor hereby grants to You a perpetual, 68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 69 | copyright license to reproduce, prepare Derivative Works of, 70 | publicly display, publicly perform, sublicense, and distribute the 71 | Work and such Derivative Works in Source or Object form. 72 | 73 | 3. Grant of Patent License. Subject to the terms and conditions of 74 | this License, each Contributor hereby grants to You a perpetual, 75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 76 | (except as stated in this section) patent license to make, have made, 77 | use, offer to sell, sell, import, and otherwise transfer the Work, 78 | where such license applies only to those patent claims licensable 79 | by such Contributor that are necessarily infringed by their 80 | Contribution(s) alone or by combination of their Contribution(s) 81 | with the Work to which such Contribution(s) was submitted. If You 82 | institute patent litigation against any entity (including a 83 | cross-claim or counterclaim in a lawsuit) alleging that the Work 84 | or a Contribution incorporated within the Work constitutes direct 85 | or contributory patent infringement, then any patent licenses 86 | granted to You under this License for that Work shall terminate 87 | as of the date such litigation is filed. 88 | 89 | 4. Redistribution. You may reproduce and distribute copies of the 90 | Work or Derivative Works thereof in any medium, with or without 91 | modifications, and in Source or Object form, provided that You 92 | meet the following conditions: 93 | 94 | (a) You must give any other recipients of the Work or 95 | Derivative Works a copy of this License; and 96 | 97 | (b) You must cause any modified files to carry prominent notices 98 | stating that You changed the files; and 99 | 100 | (c) You must retain, in the Source form of any Derivative Works 101 | that You distribute, all copyright, patent, trademark, and 102 | attribution notices from the Source form of the Work, 103 | excluding those notices that do not pertain to any part of 104 | the Derivative Works; and 105 | 106 | (d) If the Work includes a "NOTICE" text file as part of its 107 | distribution, then any Derivative Works that You distribute must 108 | include a readable copy of the attribution notices contained 109 | within such NOTICE file, excluding those notices that do not 110 | pertain to any part of the Derivative Works, in at least one 111 | of the following places: within a NOTICE text file distributed 112 | as part of the Derivative Works; within the Source form or 113 | documentation, if provided along with the Derivative Works; or, 114 | within a display generated by the Derivative Works, if and 115 | wherever such third-party notices normally appear. The contents 116 | of the NOTICE file are for informational purposes only and 117 | do not modify the License. You may add Your own attribution 118 | notices within Derivative Works that You distribute, alongside 119 | or as an addendum to the NOTICE text from the Work, provided 120 | that such additional attribution notices cannot be construed 121 | as modifying the License. 122 | 123 | You may add Your own copyright statement to Your modifications and 124 | may provide additional or different license terms and conditions 125 | for use, reproduction, or distribution of Your modifications, or 126 | for any such Derivative Works as a whole, provided Your use, 127 | reproduction, and distribution of the Work otherwise complies with 128 | the conditions stated in this License. 129 | 130 | 5. Submission of Contributions. Unless You explicitly state otherwise, 131 | any Contribution intentionally submitted for inclusion in the Work 132 | by You to the Licensor shall be under the terms and conditions of 133 | this License, without any additional terms or conditions. 134 | Notwithstanding the above, nothing herein shall supersede or modify 135 | the terms of any separate license agreement you may have executed 136 | with Licensor regarding such Contributions. 137 | 138 | 6. Trademarks. This License does not grant permission to use the trade 139 | names, trademarks, service marks, or product names of the Licensor, 140 | except as required for reasonable and customary use in describing the 141 | origin of the Work and reproducing the content of the NOTICE file. 142 | 143 | 7. Disclaimer of Warranty. Unless required by applicable law or 144 | agreed to in writing, Licensor provides the Work (and each 145 | Contributor provides its Contributions) on an "AS IS" BASIS, 146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or 147 | implied, including, without limitation, any warranties or conditions 148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A 149 | PARTICULAR PURPOSE. You are solely responsible for determining the 150 | appropriateness of using or redistributing the Work and assume any 151 | risks associated with Your exercise of permissions under this License. 152 | 153 | 8. Limitation of Liability. In no event and under no legal theory, 154 | whether in tort (including negligence), contract, or otherwise, 155 | unless required by applicable law (such as deliberate and grossly 156 | negligent acts) or agreed to in writing, shall any Contributor be 157 | liable to You for damages, including any direct, indirect, special, 158 | incidental, or consequential damages of any character arising as a 159 | result of this License or out of the use or inability to use the 160 | Work (including but not limited to damages for loss of goodwill, 161 | work stoppage, computer failure or malfunction, or any and all 162 | other commercial damages or losses), even if such Contributor 163 | has been advised of the possibility of such damages. 164 | 165 | 9. Accepting Warranty or Additional Liability. While redistributing 166 | the Work or Derivative Works thereof, You may choose to offer, 167 | and charge a fee for, acceptance of support, warranty, indemnity, 168 | or other liability obligations and/or rights consistent with this 169 | License. However, in accepting such obligations, You may act only 170 | on Your own behalf and on Your sole responsibility, not on behalf 171 | of any other Contributor, and only if You agree to indemnify, 172 | defend, and hold each Contributor harmless for any liability 173 | incurred by, or claims asserted against, such Contributor by reason 174 | of your accepting any such warranty or additional liability. 175 | 176 | END OF TERMS AND CONDITIONS 177 | 178 | APPENDIX: How to apply the Apache License to your work. 179 | 180 | To apply the Apache License to your work, attach the following 181 | boilerplate notice, with the fields enclosed by brackets "[]" 182 | replaced with your own identifying information. (Don't include 183 | the brackets!) The text should be enclosed in the appropriate 184 | comment syntax for the file format. We also recommend that a 185 | file or class name and description of purpose be included on the 186 | same "printed page" as the copyright notice for easier 187 | identification within third-party archives. 188 | 189 | Copyright [yyyy] [name of copyright owner] 190 | 191 | Licensed under the Apache License, Version 2.0 (the "License"); 192 | you may not use this file except in compliance with the License. 193 | You may obtain a copy of the License at 194 | 195 | http://www.apache.org/licenses/LICENSE-2.0 196 | 197 | Unless required by applicable law or agreed to in writing, software 198 | distributed under the License is distributed on an "AS IS" BASIS, 199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 200 | See the License for the specific language governing permissions and 201 | limitations under the License. 202 | -------------------------------------------------------------------------------- /MANIFEST.in: -------------------------------------------------------------------------------- 1 | include LICENSE 2 | prune tests 3 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 |

sympytorch

2 | 3 | Turn SymPy expressions into PyTorch Modules. 4 | 5 | SymPy floats (optionally) become trainable parameters. SymPy symbols are inputs to the Module. 6 | 7 | Optimise your symbolic expressions via gradient descent! 8 | 9 | ## Installation 10 | 11 | ```bash 12 | pip install sympytorch 13 | ``` 14 | Requires Python 3.7+ and PyTorch 1.6.0+ and SymPy 1.7.1+. 15 | 16 | ## Example 17 | 18 | ```python 19 | import sympy, torch, sympytorch 20 | 21 | x = sympy.symbols('x_name') 22 | cosx = 1.0 * sympy.cos(x) 23 | sinx = 2.0 * sympy.sin(x) 24 | mod = sympytorch.SymPyModule(expressions=[cosx, sinx]) 25 | 26 | x_ = torch.rand(3) 27 | out = mod(x_name=x_) # out has shape (3, 2) 28 | 29 | assert torch.equal(out[:, 0], x_.cos()) 30 | assert torch.equal(out[:, 1], 2 * x_.sin()) 31 | assert out.requires_grad # from the two Parameters initialised as 1.0 and 2.0 32 | assert {x.item() for x in mod.parameters()} == {1.0, 2.0} 33 | ``` 34 | 35 | ## API 36 | 37 | ```python 38 | sympytorch.SymPyModule(*, expressions, extra_funcs=None) 39 | ``` 40 | Where: 41 | - `expressions` is a list of SymPy expressions. 42 | - `extra_funcs` is a dictionary mapping from custom `sympy.Function`s to their PyTorch implementation. Defaults to no extra functions. 43 | 44 | Instances of `SymPyModule` can be called, passing the values of the symbols as in the above example. 45 | 46 | `SymPyModule` has a method `.sympy()`, which returns the corresponding list of SymPy expressions. (Which may not be the same as the expressions it was initialised with, if the values of its Parameters have been changed, i.e. have been learnt.) 47 | 48 | Wrapping floats in `sympy.UnevaluatedExpr` will cause them not to be trained, by registering them as buffers rather than parameters. 49 | 50 | ```python 51 | sympytorch.hide_floats(expression) 52 | ``` 53 | As a convenience, `hide_floats` will take an expression and return a new expression with every float wrapped in a `sympy.UnevaluatedExpr`, so that it is interpreted as a buffer rather than a parameter. 54 | 55 | ## Extensions 56 | 57 | Not every PyTorch or SymPy operation is supported -- just the ones that I found I've needed! There's a dictionary [here](./sympytorch/sympy_module.py#L12) that lists the supported operations. Feel free to submit PRs for any extra operations you think should be in by default. You can also use the `extra_funcs` argument to specify extra functions, including custom functions. 58 | -------------------------------------------------------------------------------- /setup.py: -------------------------------------------------------------------------------- 1 | import pathlib 2 | import re 3 | 4 | import setuptools 5 | 6 | 7 | _here = pathlib.Path(__file__).resolve().parent 8 | 9 | 10 | name = "sympytorch" 11 | 12 | # for simplicity we actually store the version in the __version__ attribute in the 13 | # source 14 | with open(_here / name / "__init__.py") as f: 15 | meta_match = re.search(r"^__version__ = ['\"]([^'\"]*)['\"]", f.read(), re.M) 16 | if meta_match: 17 | version = meta_match.group(1) 18 | else: 19 | raise RuntimeError("Unable to find __version__ string.") 20 | 21 | author = "Patrick Kidger" 22 | 23 | author_email = "contact@kidger.site" 24 | 25 | description = "Turning SymPy expressions into PyTorch modules." 26 | 27 | with open(_here / "README.md", "r") as f: 28 | readme = f.read() 29 | 30 | url = "https://github.com/patrick-kidger/sympytorch" 31 | 32 | license = "Apache-2.0" 33 | 34 | classifiers = [ 35 | "Development Status :: 3 - Alpha", 36 | "Intended Audience :: Developers", 37 | "Intended Audience :: Financial and Insurance Industry", 38 | "Intended Audience :: Information Technology", 39 | "Intended Audience :: Science/Research", 40 | "License :: OSI Approved :: Apache Software License", 41 | "Natural Language :: English", 42 | "Programming Language :: Python :: 3", 43 | "Topic :: Scientific/Engineering :: Artificial Intelligence", 44 | "Topic :: Scientific/Engineering :: Information Analysis", 45 | "Topic :: Scientific/Engineering :: Mathematics", 46 | ] 47 | 48 | python_requires = "~=3.7" 49 | 50 | install_requires = ["torch>=1.6.0", "sympy>=1.7.1"] 51 | 52 | setuptools.setup( 53 | name=name, 54 | version=version, 55 | author=author, 56 | author_email=author_email, 57 | maintainer=author, 58 | maintainer_email=author_email, 59 | description=description, 60 | long_description=readme, 61 | long_description_content_type="text/markdown", 62 | url=url, 63 | license=license, 64 | classifiers=classifiers, 65 | zip_safe=False, 66 | python_requires=python_requires, 67 | install_requires=install_requires, 68 | packages=[name], 69 | package_data={name: ["py.typed"]}, 70 | ) 71 | -------------------------------------------------------------------------------- /sympytorch/__init__.py: -------------------------------------------------------------------------------- 1 | from .hide_floats_m import hide_floats as hide_floats 2 | from .sympy_module import SymPyModule as SymPyModule 3 | 4 | 5 | __version__ = "0.1.4" 6 | -------------------------------------------------------------------------------- /sympytorch/hide_floats_m.py: -------------------------------------------------------------------------------- 1 | from typing import Dict, TypeVar 2 | 3 | import sympy 4 | 5 | 6 | ExprType = TypeVar("ExprType", bound=sympy.Expr) 7 | 8 | 9 | def hide_floats(expression: ExprType) -> ExprType: 10 | _memodict: Dict[ExprType, ExprType] = {} 11 | return _hide_floats(expression, _memodict) 12 | 13 | 14 | def _hide_floats(expression: ExprType, _memodict: Dict[ExprType, ExprType]) -> ExprType: 15 | try: 16 | return _memodict[expression] 17 | except KeyError: 18 | pass 19 | 20 | if issubclass(expression.func, (sympy.Min, sympy.Max)): 21 | evaluate = False 22 | else: 23 | evaluate = True 24 | 25 | new_expression: ExprType 26 | if issubclass(expression.func, sympy.Float): 27 | new_expression = sympy.UnevaluatedExpr(expression) # type: ignore 28 | elif issubclass(expression.func, sympy.Integer): 29 | new_expression = expression 30 | elif issubclass(expression.func, sympy.Symbol): 31 | new_expression = expression 32 | else: 33 | new_expression = expression.func( 34 | *[_hide_floats(arg, _memodict) for arg in expression.args], 35 | evaluate=evaluate 36 | ) # type: ignore 37 | _memodict[expression] = new_expression 38 | return new_expression 39 | -------------------------------------------------------------------------------- /sympytorch/py.typed: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/patrick-kidger/sympytorch/0670a3bbdfc0b77fa1ecc1367f51887eb6c6e3e9/sympytorch/py.typed -------------------------------------------------------------------------------- /sympytorch/sympy_module.py: -------------------------------------------------------------------------------- 1 | from __future__ import annotations 2 | 3 | import collections as co 4 | import functools as ft 5 | from typing import ( 6 | Any, 7 | Callable, 8 | Dict, 9 | Generic, 10 | List, 11 | Sequence, 12 | Tuple, 13 | Type, 14 | TYPE_CHECKING, 15 | TypeVar, 16 | Union, 17 | ) 18 | 19 | import sympy 20 | import torch 21 | 22 | 23 | ExprType = TypeVar("ExprType", bound=sympy.Expr) 24 | T = TypeVar("T") 25 | 26 | if TYPE_CHECKING: 27 | # Because there are methods of our class objects below called `sympy` that 28 | # implicitly override the `sympy` name in the global namespace while defining other 29 | # methods, our type checker needs to know that we're referring to the sympy module 30 | # in our type annotations. 31 | import sympy as sympy_ 32 | 33 | 34 | def _reduce(fn: Callable[..., T]) -> Callable[..., T]: 35 | def fn_(*args: Any) -> T: 36 | return ft.reduce(fn, args) 37 | 38 | return fn_ 39 | 40 | 41 | def _I(*args: Any) -> torch.Tensor: 42 | return torch.tensor(1j) 43 | 44 | 45 | _global_func_lookup: Dict[ 46 | Union[Type[sympy.Basic], Callable[..., Any]], Callable[..., torch.Tensor] 47 | ] = { 48 | sympy.Mul: _reduce(torch.mul), 49 | sympy.Add: _reduce(torch.add), 50 | sympy.div: torch.div, 51 | sympy.Abs: torch.abs, 52 | sympy.sign: torch.sign, 53 | # Note: May raise error for ints. 54 | sympy.ceiling: torch.ceil, 55 | sympy.floor: torch.floor, 56 | sympy.log: torch.log, 57 | sympy.exp: torch.exp, 58 | sympy.sqrt: torch.sqrt, 59 | sympy.cos: torch.cos, 60 | sympy.acos: torch.acos, 61 | sympy.sin: torch.sin, 62 | sympy.asin: torch.asin, 63 | sympy.tan: torch.tan, 64 | sympy.atan: torch.atan, 65 | sympy.atan2: torch.atan2, 66 | # Note: May give NaN for complex results. 67 | sympy.cosh: torch.cosh, 68 | sympy.acosh: torch.acosh, 69 | sympy.sinh: torch.sinh, 70 | sympy.asinh: torch.asinh, 71 | sympy.tanh: torch.tanh, 72 | sympy.atanh: torch.atanh, 73 | sympy.Pow: torch.pow, 74 | sympy.re: torch.real, 75 | sympy.im: torch.imag, 76 | sympy.arg: torch.angle, 77 | # Note: May raise error for ints and complexes 78 | sympy.erf: torch.erf, 79 | sympy.loggamma: torch.lgamma, 80 | sympy.Eq: torch.eq, 81 | sympy.Ne: torch.ne, 82 | sympy.StrictGreaterThan: torch.gt, 83 | sympy.StrictLessThan: torch.lt, 84 | sympy.LessThan: torch.le, 85 | sympy.GreaterThan: torch.ge, 86 | sympy.And: torch.logical_and, 87 | sympy.Or: torch.logical_or, 88 | sympy.Not: torch.logical_not, 89 | sympy.Max: torch.max, 90 | sympy.Min: torch.min, 91 | # Matrices 92 | sympy.MatAdd: torch.add, 93 | sympy.HadamardProduct: torch.mul, 94 | sympy.Trace: torch.trace, 95 | # Note: May raise error for integer matrices. 96 | sympy.Determinant: torch.det, 97 | sympy.core.numbers.ImaginaryUnit: _I, 98 | sympy.conjugate: torch.conj, 99 | } 100 | 101 | number_symbols = [cls for cls in sympy.NumberSymbol.__subclasses__()] 102 | 103 | 104 | def number_symbol_to_torch(symbol: sympy.NumberSymbol, *args: Any) -> torch.Tensor: 105 | return torch.tensor(float(symbol)) 106 | 107 | 108 | _global_func_lookup.update( 109 | {s: ft.partial(number_symbol_to_torch, s()) for s in number_symbols} 110 | ) 111 | 112 | 113 | class _Node(torch.nn.Module, Generic[ExprType]): 114 | def __init__( 115 | self, 116 | *, 117 | expr: ExprType, 118 | _memodict: Dict[sympy.Basic, torch.nn.Module], 119 | _func_lookup, 120 | **kwargs, 121 | ) -> None: 122 | super().__init__(**kwargs) 123 | 124 | self._sympy_func: Type[ExprType] = expr.func 125 | 126 | self._torch_func: Callable[..., torch.Tensor] 127 | self._args: Union[ 128 | torch.nn.ModuleList, 129 | Tuple[Callable[[Dict[str, torch.Tensor]], torch.Tensor], ...], 130 | ] 131 | self._value: Any 132 | 133 | if issubclass(expr.func, sympy.Float): 134 | self._value = torch.nn.Parameter(torch.tensor(float(expr))) 135 | self._torch_func = lambda: self._value 136 | self._args = () 137 | elif issubclass(expr.func, sympy.Integer): 138 | self._value = torch.tensor(int(expr)) 139 | self._torch_func = lambda: self._value 140 | self._args = () 141 | elif issubclass(expr.func, sympy.Rational): 142 | self._numerator: torch.Tensor 143 | self._denominator: torch.Tensor 144 | assert isinstance(expr, sympy.Rational) 145 | self.register_buffer( 146 | "_numerator", torch.tensor(expr.p, dtype=torch.get_default_dtype()) 147 | ) 148 | self.register_buffer( 149 | "_denominator", torch.tensor(expr.q, dtype=torch.get_default_dtype()) 150 | ) 151 | self._torch_func = lambda: self._numerator / self._denominator 152 | self._args = () 153 | elif issubclass(expr.func, sympy.UnevaluatedExpr): 154 | if len(expr.args) != 1 or not issubclass(expr.args[0].func, sympy.Float): 155 | raise ValueError("UnevaluatedExpr should only be used to wrap floats.") 156 | assert isinstance(expr.args[0], sympy.Float) 157 | self.register_buffer("_value", torch.tensor(float(expr.args[0]))) 158 | self._torch_func = lambda: self._value 159 | self._args = () 160 | elif issubclass(expr.func, sympy.Symbol): 161 | assert isinstance(expr, sympy.Symbol) 162 | self._name = expr.name 163 | self._torch_func = lambda value: value 164 | self._args = ((lambda memodict: memodict[expr.name]),) 165 | else: 166 | self._torch_func = _func_lookup[expr.func] 167 | args: List[torch.nn.Module] = [] 168 | for arg in expr.args: 169 | try: 170 | arg_ = _memodict[arg] 171 | except KeyError: 172 | arg_ = type(self)( 173 | expr=arg, # type: ignore 174 | _memodict=_memodict, 175 | _func_lookup=_func_lookup, 176 | **kwargs, 177 | ) 178 | _memodict[arg] = arg_ 179 | args.append(arg_) 180 | self._args = torch.nn.ModuleList(args) 181 | 182 | def sympy(self, _memodict: Dict[_Node, sympy_.Expr]) -> ExprType: 183 | if issubclass(self._sympy_func, sympy.Float): 184 | assert isinstance(self._value, torch.nn.Parameter) 185 | return self._sympy_func(self._value.item()) 186 | elif issubclass(self._sympy_func, sympy.UnevaluatedExpr): 187 | assert isinstance(self._value, torch.Tensor) 188 | return self._sympy_func(self._value.item()) 189 | elif issubclass( 190 | self._sympy_func, 191 | (type(sympy.S.NegativeOne), type(sympy.S.One), type(sympy.S.Zero)), 192 | ): 193 | return self._sympy_func() 194 | elif issubclass(self._sympy_func, sympy.Integer): 195 | return self._sympy_func(self._value) 196 | elif issubclass(self._sympy_func, sympy.Rational): 197 | if issubclass(self._sympy_func, type(sympy.S.Half)): 198 | return sympy.S.Half 199 | else: 200 | return self._sympy_func( 201 | self._numerator.item(), self._denominator.item() 202 | ) 203 | elif issubclass(self._sympy_func, sympy.Symbol): 204 | return self._sympy_func(self._name) 205 | elif issubclass(self._sympy_func, sympy.core.numbers.ImaginaryUnit): 206 | return sympy.I 207 | elif issubclass(self._sympy_func, sympy.core.numbers.NumberSymbol): 208 | return self._sympy_func() 209 | else: 210 | if issubclass(self._sympy_func, (sympy.Min, sympy.Max)): 211 | evaluate = False 212 | else: 213 | evaluate = True 214 | args = [] 215 | for arg in self._args: 216 | assert isinstance(arg, _Node) 217 | try: 218 | arg_ = _memodict[arg] 219 | except KeyError: 220 | arg_ = arg.sympy(_memodict) 221 | _memodict[arg] = arg_ 222 | args.append(arg_) 223 | return self._sympy_func(*args, evaluate=evaluate) # type: ignore 224 | 225 | def forward(self, memodict) -> torch.Tensor: 226 | args = [] 227 | for arg in self._args: 228 | try: 229 | arg_ = memodict[arg] 230 | except KeyError: 231 | arg_ = arg(memodict) 232 | memodict[arg] = arg_ 233 | args.append(arg_) 234 | return self._torch_func(*args) 235 | 236 | 237 | class SymPyModule(torch.nn.Module): 238 | def __init__(self, *, expressions, extra_funcs=None, **kwargs): 239 | super().__init__(**kwargs) 240 | 241 | expressions = tuple(expressions) 242 | 243 | if extra_funcs is None: 244 | extra_funcs = {} 245 | _func_lookup = co.ChainMap(_global_func_lookup, extra_funcs) 246 | 247 | _memodict = {} 248 | self._nodes: Sequence[_Node] = torch.nn.ModuleList( # type: ignore 249 | [ 250 | _Node(expr=expr, _memodict=_memodict, _func_lookup=_func_lookup) 251 | for expr in expressions 252 | ] 253 | ) 254 | self._expressions_string = str(expressions) 255 | 256 | def __repr__(self): 257 | return f"{type(self).__name__}(expressions={self._expressions_string})" 258 | 259 | def sympy(self) -> List[sympy.Expr]: 260 | _memodict: Dict[_Node, sympy.Expr] = {} 261 | return [node.sympy(_memodict) for node in self._nodes] 262 | 263 | def forward(self, **symbols: Any) -> torch.Tensor: 264 | out = [node(symbols) for node in self._nodes] 265 | out = torch.broadcast_tensors(*out) 266 | return torch.stack(out, dim=-1) 267 | -------------------------------------------------------------------------------- /tests/test_hide_floats.py: -------------------------------------------------------------------------------- 1 | import sympy 2 | 3 | import sympytorch 4 | 5 | 6 | def test_hide_floats(): 7 | x = sympy.Symbol("x") 8 | y = sympy.Symbol("y") 9 | e = sympy.UnevaluatedExpr 10 | 11 | z = 3.0 * x + (4.0 * y + 1.0) ** 2 - 1 12 | z_ = sympytorch.hide_floats(z) 13 | assert z_ == x * e(3.0) - 1 + e(16.0) * (y + e(0.25)) ** 2 14 | -------------------------------------------------------------------------------- /tests/test_sympymodule.py: -------------------------------------------------------------------------------- 1 | import numpy as np 2 | import sympy 3 | import torch 4 | 5 | import sympytorch 6 | 7 | 8 | def test_example(): 9 | x = sympy.symbols("x_name") 10 | cosx = 1.0 * sympy.cos(x) 11 | sinx = 2.0 * sympy.sin(x) 12 | 13 | mod = sympytorch.SymPyModule(expressions=[cosx, sinx]) 14 | 15 | x_ = torch.rand(3) 16 | out = mod(x_name=x_) # out has shape (3, 2) 17 | 18 | assert torch.equal(out[:, 0], x_.cos()) 19 | assert torch.equal(out[:, 1], 2 * x_.sin()) 20 | assert out.requires_grad # from the two Parameters initialised as 1.0 and 2.0 21 | assert {x.item() for x in mod.parameters()} == {1.0, 2.0} 22 | 23 | 24 | def test_grad(): 25 | x = sympy.symbols("x_name") 26 | y = 1.0 * x 27 | mod = sympytorch.SymPyModule(expressions=[y]) 28 | out = mod(x_name=torch.ones(())) 29 | out.backward() 30 | with torch.no_grad(): 31 | for param in mod.parameters(): 32 | param += param.grad 33 | (expr,) = mod.sympy() 34 | assert expr == 2.0 * x 35 | 36 | 37 | def test_reduce(): 38 | x, y = sympy.symbols("x y") 39 | z = 2 * x * y 40 | mod = sympytorch.SymPyModule(expressions=[z]) 41 | mod(x=torch.rand(2), y=torch.rand(2)) 42 | 43 | z = 2 + x + y 44 | mod = sympytorch.SymPyModule(expressions=[z]) 45 | mod(x=torch.rand(2), y=torch.rand(2)) 46 | 47 | 48 | def test_special_subclasses(): 49 | x, y = sympy.symbols("x y") 50 | z = x - 1 51 | w = y * 0 52 | u = sympy.Integer(1) 53 | 54 | mod = sympytorch.SymPyModule(expressions=[z, w, u]) 55 | assert mod.sympy() == [x - 1, sympy.Integer(0), sympy.Integer(1)] 56 | 57 | 58 | def test_constants(): 59 | x = sympy.symbols("x") 60 | y = 2.0 * x + sympy.UnevaluatedExpr(1.0) 61 | mod = sympytorch.SymPyModule(expressions=[y]) 62 | assert mod.sympy() == [y] 63 | assert set(p.item() for p in mod.parameters()) == {2.0} 64 | assert set(b.item() for b in mod.buffers()) == {1.0} 65 | 66 | 67 | def test_custom_function(): 68 | x, y = sympy.symbols("x y") 69 | f = sympy.Function("f") 70 | z = x + f(y) 71 | extra_funcs = {f: lambda y_: y_**2} 72 | mod = sympytorch.SymPyModule(expressions=[z], extra_funcs=extra_funcs) 73 | assert mod.sympy() == [z] 74 | assert mod(x=1, y=2) == 1 + 2**2 75 | 76 | 77 | def test_rationals(): 78 | xvals = np.random.randn(100) 79 | x = sympy.symbols("x") 80 | y = x * sympy.Rational(2, 7) 81 | mod = sympytorch.SymPyModule(expressions=[y]) 82 | mod.to(torch.float64) 83 | assert mod.sympy() == [y], "mod: {}, y:{}".format(mod.sympy(), y) 84 | assert len([p.item() for p in mod.parameters()]) == 0 85 | y_tilde = mod(x=torch.tensor(xvals, dtype=torch.float64))[:, 0] 86 | error = y_tilde.detach().numpy() - xvals * 2 / 7 87 | assert (error**2).mean() < 1e-10, "error:{}".format((error**2).mean()) 88 | 89 | 90 | def test_half1(): 91 | xvals = np.random.randn(100) 92 | x = sympy.symbols("x") 93 | y = abs(x) ** sympy.S.Half 94 | mod = sympytorch.SymPyModule(expressions=[y]) 95 | mod.to(torch.float64) 96 | assert mod.sympy() == [y], "mod: {}, y:{}".format(mod.sympy(), y) 97 | y_tilde = mod(x=torch.tensor(xvals, dtype=torch.float64))[:, 0] 98 | error = y_tilde.detach().numpy() - np.abs(xvals) ** 0.5 99 | assert (error**2).mean() < 1e-10, "error:{}".format((error**2).mean()) 100 | 101 | 102 | def test_half2(): 103 | xvals = np.random.randn(100) 104 | y = sympy.parse_expr("sqrt(Abs(x))") 105 | mod = sympytorch.SymPyModule(expressions=[y]) 106 | mod.to(torch.float64) 107 | assert mod.sympy() == [y], "mod: {}, y:{}".format(mod.sympy(), y) 108 | assert len([p.item() for p in mod.parameters()]) == 0 109 | y_tilde = mod(x=torch.tensor(xvals, dtype=torch.float64))[:, 0] 110 | error = y_tilde.detach().numpy() - np.abs(xvals) ** 0.5 111 | assert (error**2).mean() < 1e-10, "error:{}".format((error**2).mean()) 112 | 113 | 114 | def test_constants2(): 115 | constants = [ 116 | sympy.pi, 117 | sympy.E, 118 | sympy.GoldenRatio, 119 | sympy.TribonacciConstant, 120 | sympy.EulerGamma, 121 | sympy.Catalan, 122 | ] 123 | mod = sympytorch.SymPyModule(expressions=constants) 124 | mod.to(torch.float64) 125 | assert mod.sympy() == constants, "mod: {}, y:{}".format(mod.sympy(), constants) 126 | assert len([p.item() for p in mod.parameters()]) == 0 127 | torch.testing.assert_allclose(mod(), torch.tensor([float(c) for c in constants])) 128 | 129 | 130 | def test_complex(): 131 | # Simple complex number handing test 132 | x = sympy.symbols("x") 133 | 134 | complex_func_torch = sympytorch.SymPyModule( 135 | expressions=[ 136 | x * sympy.I, 137 | sympy.conjugate(x), 138 | sympy.sqrt(sympy.conjugate(x * sympy.I) * x * sympy.I), 139 | ] 140 | ) 141 | 142 | out = complex_func_torch(x=torch.tensor(2.0, dtype=torch.double)).detach().numpy() 143 | assert out[0].item() == 2.0j, "Expected 2j, eval:{}".format(out[0].item()) 144 | assert out[1].item() == 2.0, "Expected 2, eval:{}".format(out[1].item()) 145 | assert out[2].item() == 2.0, "Expected 2, eval:{}".format(out[2].item()) 146 | 147 | # Complex number handling test with complex parameters 148 | out = ( 149 | complex_func_torch(x=torch.tensor(2.0j, dtype=torch.complex128)) 150 | .detach() 151 | .numpy() 152 | ) 153 | assert out[0].item() == -2.0, "Expected -2, eval:{}".format(out[0].item()) 154 | assert out[1].item() == -2.0j, "Expected -2j, eval:{}".format(out[1].item()) 155 | assert out[2].item() == 2.0, "Expected 2, eval:{}".format(out[2].item()) 156 | 157 | # Comparison with numpy using spherical harmonics 158 | theta, phi = sympy.symbols("theta phi") 159 | max_l = 2 160 | m_list = range(-max_l, max_l + 1) 161 | 162 | # spherical harmonics from l=-2 to l=2 163 | func_list = [ 164 | sympy.simplify( 165 | sympy.functions.special.spherical_harmonics.Znm( 166 | max_l, m, theta, phi 167 | ).expand(func=True) 168 | ).evalf() 169 | for m in m_list 170 | ] 171 | 172 | # Numpy and Torch based functions 173 | func_list_np = [sympy.lambdify([theta, phi], func, "numpy") for func in func_list] 174 | func_list_torch = sympytorch.SymPyModule(expressions=func_list) 175 | 176 | np_eval = np.array( 177 | list(map(lambda i: func_list_np[i](np.pi / 3, np.pi / 3), np.arange(5))) 178 | ) 179 | torch_eval = func_list_torch( 180 | theta=torch.tensor(np.pi / 3, dtype=torch.double), 181 | phi=torch.tensor(np.pi / 3, dtype=torch.double), 182 | ) 183 | 184 | # Correctness within single precision 185 | error = np.sum(np.abs(torch_eval.detach().numpy() - np_eval)) 186 | assert error < 1e-7, "np v torch complex error:{}".format(error) 187 | 188 | 189 | def test_integers(): 190 | m = sympytorch.SymPyModule(expressions=[sympy.core.numbers.Zero()]) 191 | assert m() == torch.tensor([0.0]) 192 | 193 | m = sympytorch.SymPyModule(expressions=[sympy.core.numbers.One()]) 194 | assert m() == torch.tensor([1.0]) 195 | 196 | m = sympytorch.SymPyModule(expressions=[sympy.core.numbers.NegativeOne()]) 197 | assert m() == torch.tensor([-1.0]) 198 | 199 | for i in range(-10, 10): 200 | m = sympytorch.SymPyModule(expressions=[sympy.core.numbers.Integer(i)]) 201 | assert m() == torch.tensor([float(i)]) 202 | --------------------------------------------------------------------------------