pytensor
pytensor copied to clipboard
pymc-devs
PyTensor allows you to define, optimize, and efficiently evaluate mathematical expressions involving multi-dimensional arrays.
### Description In PyMC we have a SymbolicRandomVariable that allows defining an OpFromGraph that looks almost like a RandomVariable but is just built symbolically. This avoids having to define too...
### Description With numba 0.61 coverage will work for lines that are seen during compilation: https://numba.readthedocs.io/en/stable/release/0.61.0-notes.html#added-compile-time-code-coverage This means we may be able to stop running tests in object mode just...
## Description Add a decorator that transforms a JAX function such that it can be used in PyTensor. Shape and dtype inference works automatically and input and output can be...
### Description If we follow https://scientific-python.org/specs/spec-0000/ numpy < 2 should be dropped around September 2025.
> I've noticed that some of the code like this is redundant now. E.g. `x * np.array(math.log(2.0)).astype(x.dtype)` is now the same as `x * math.log(2.0)` because numpy now uses the...
### Description We did some work recently to cut down PyTensor import time. We should protect against unnoticed regressions by testing explicitly. Not sure this can be done inside a...
### Description We have very long modules with arbitrary names like `tensor.basic` vs `tensor.extra_ops`, and `nlinalg` vs `slinalg`. We should organize the modules by operator functionality. I would error on...
### Description All general Blockwise methods (`infer_shape`, `grad`, ...) work for Elemwise which is just a trivial Blockwise function with signature "(),()...()->(),...,()` (that is one entry for each input and...
## Description Implement scipy optimization routines, with implicit gradients. This PR should add: - [ ] `optimize.minimize` - [ ] `optimize.root` - [ ] `optimize.scalar_minimize` - [ ] `optimize.scalar_root` It...
## Description Replaced all instances of `output_type_depends_on_input_value` with `data_dependent_output_shapes` for discoverability / standardization. Follows the Aarray API standard for this: https://data-apis.org/array-api/draft/design_topics/data_dependent_output_shapes.html ## Related Issue - [x] Closes #1188 - [...
