Releases: aesara-devs/aesara
Releases · aesara-devs/aesara
rel-2.8.7
What's Changed
- Add deprecation warning for
rng_mrg
by @rlouf in #1187 - Fix
None
output node bug inJITLinker.create_jitable_thunk
by @ricardoV94 in #1205 - Update
pre-commit
hooks by @brandonwillard in #1208 - Make
isort
ignore Cython files by @brandonwillard in #1209 - Update Scan Gibbs sampling example by @LegrandNico in #1189
- Pin style CI job to Python 3.9 by @brandonwillard in #1210
- Fix outdated stride logic in
GEMM
's C implementation by @brandonwillard in #1219 - Stop renaming variables in
Scan
's merge rewrite by @brandonwillard in #1218 - Allow shared
RandomState
/Generator
updates in JAX backend by @ricardoV94 in #1214 - Fix
_debugprint
's handling of empty profile data by @brandonwillard in #1224 - Fix storage and mit-mot handling in
numba_funcify_Scan
by @brandonwillard in #1203 - Make
fgraph_to_python
processConstant
FunctionGraph
outputs correctly by @brandonwillard in #1234
Full Changelog: rel-2.8.6...rel-2.8.7
rel-2.8.6
What's Changed
- Fix type inference in Elemwise when inputs have 0 shape by @ricardoV94 in #1199
- Fix split output type by @ricardoV94 in #1201
- Avoid duplicate input names in Numba's
Scan
transpilation by @rlouf in #1198 - Implement moveaxis helper by @ricardoV94 in #1186
Full Changelog: rel-2.8.5...rel-2.8.6
rel-2.8.5
What's Changed
- Remove JAX omnistaging warning and add a warning for
CheckAndRaise
transpilation by @rlouf in #1185 - Update
CheckAndRaise
to handleScalarType
inputs by @brandonwillard in #1181 - Clean up
SparseTensorType
by @brandonwillard in #1192 - Allow
TensorType.convert_variable
to work with mixed static shapes by @ricardoV94 in #1190
Full Changelog: rel-2.8.4...rel-2.8.5
rel-2.8.4
rel-2.8.3
What's Changed
- Refactor setuptools to use declarative
setup.cfg
by @maresb in #1135 - Remove defaults channel from conda env for docs by @maresb in #1136
- Split
aesara.tensor.rewriting.basic
rewrites by theiraesara.tensor
modules by @brandonwillard in #1139 - Move
tensor_copy
rewrite toaesara.tensor.rewriting.basic
by @brandonwillard in #1144 - Add docstrings for
RandomVariable
Op
s by @rlouf in #1105 - Support duplicate indices in
AdvancedIncSubtensor1
's Numba implementation by @brandonwillard in #1143 - Add a C implementation for
BroadcastTo
by @brandonwillard in #1130 - Move Numba tests to their own sub-package by @brandonwillard in #1153
- Add a rewrite that changes constant indices to unsigned integers by @brandonwillard in #1150
- Add tensor conversion to Numpy-like helper functions by @ltoniazzi in #1154
- Make
ChoiceRV
behave like NumPy'schoice
by @rlouf in #1163 - Add tensor conversion to
flatnonzero
,nonzero_values
,tile
,inverse_permutation
, anddiag
by @ltoniazzi in #1166 - Use static shape in outputs of
Elemwise
by @ricardoV94 in #821 - Refactor the JAX dispatcher by @rlouf in #1168
New Contributors
- @ltoniazzi made their first contribution in #1154
Full Changelog: rel-2.8.2...rel-2.8.3
rel-2.8.2
What's Changed
- Fix broadcasting conditions in
broadcast_shape_iter
by @brandonwillard in #1138
Full Changelog: rel-2.8.1...rel-2.8.2
rel-2.8.1
What's Changed
- Fix handling of
numpy.dtype
input inTensorType
by @brandonwillard in #1133
Full Changelog: rel-2.8.0...rel-2.8.1
rel-2.8.0
What's Changed
- Change use of "optimize" to "rewrite" by @brandonwillard in #1054
- Add a C implementation for
TensorFromScalar
by @brandonwillard in #1124 - Remove warnings from
tests.tensor.test_[basic|var]
by @brandonwillard in #1127 - Add pip as an explicit docs dependency by @maresb in #1132
- Revert to lazy loaded
aesara.link.c.cutils
by @brandonwillard in #1131
Full Changelog: rel-2.7.10...rel-2.8.0
rel-2.7.10
What's Changed
- Change R-operator to L-operator in
Lop
docstring by @brandonwillard in #1087 - Implement
aesara.tensor.matmul
by @zoj613 in #744 - Add a JAX implementation for
FillDiagonal
by @ferrine in #1080 - Add docstring for the
GammaRV
Op by @rlouf in #1084 - Refactor and fix static shape issues in
IfElse
by @brandonwillard in #1095 - Add docstrings for
RandomVariable
s by @rlouf in #1091 - Fix incorrect setuptools minimum version number by @maresb in #1100
- Add docstrings for dirichlet, geometric and hypergeometric distributions by @rlouf in #1099
- Allow
size
to broadcastCategoricalRV
p
argument by @ricardoV94 in #1088 - Add Owens-T Function
Op
by @RuneDominik in #1083 - Fix cache versioning of empty
Op.__props__
by @brandonwillard in #1108 - Allow JAX Reshape to work with constant shape inputs by @ricardoV94 in #1111
- Remove strict
TensorType.broadcastable
usage fromlocal_elemwise_alloc
by @brandonwillard in #1102 - Remove unnecessary code in
local_useless_AdvancedSubtensor1
and refactor tests by @brandonwillard in #1112
New Contributors
- @maresb made their first contribution in #1100
- @RuneDominik made their first contribution in #1083
Full Changelog: rel-2.7.9...rel-2.7.10
rel-2.7.9
What's Changed
- Fix typing error in numba MakeVector impl by @aseyboldt in #1070
- Prevent
Print
from being constant folded by @brandonwillard in #1071 - Convert Numba scalar outputs to
ndarray
s by @brandonwillard in #1073 - Add some more Numba special functions by @aseyboldt in #1077
Full Changelog: rel-2.7.8...rel-2.7.9