rel-2.28.3

What's Changed

Maintenance 🔧

• Allow transposed argument in linalg.solve by @jessegrabowski in #1231
• Do not introduce 0 strides for broadcastable dimensions in DimShuffle by @ricardoV94 in #1267

Full Changelog: rel-2.28.2...rel-2.28.3

rel-2.28.2

What's Changed

New Features 🎉

• Allow ignoring specific rewrites from optimizer_verbose by @ricardoV94 in #1240
• Allow passing trust_input to function by @Aarsh-Wankar in #1206
• Support more cases of numba advanced indexing by @ricardoV94 in #1254

Bugfixes 🐛

• Fix inner graph inplace rewrites in Numba / PyTorch backends by @ricardoV94 in #1247

Documentation 📖

• Improve elemwise docstrings by @fonnesbeck in #1255

Maintenance 🔧

• Don't use numba objectmode with vector boolean inc_subtensor by @ricardoV94 in #1243
• Speedup DimShuffle and Reshape in C and Numba backends by @ricardoV94 in #1226
• Handle MvNormal method in Op call by @ricardoV94 in #1252

Full Changelog: rel-2.28.1...rel-2.28.2

rel-2.28.1

What's Changed

New Features 🎉

• Add print_shape and print_memory_map option to debugprint by @Aarsh-Wankar in #1236

Bugfixes 🐛

• Fix solve_triangular output when overwrite_b=True by @jessegrabowski in #1235
• Fix bug in local_reshape_to_dimshuffle by @ricardoV94 in #1242

Maintenance 🔧

• Fix warning message since m2w64-toolchain is no longer installable by @maresb in #1238

New Contributors

• @Aarsh-Wankar made their first contribution in #1236

Full Changelog: rel-2.28.0...rel-2.28.1

rel-2.28.0

What's Changed

Major Changes 🛠

• Support Python 3.13 by @maresb in #1199
• Make PyTensor compatible with numpy 2.0 by @brendan-m-murphy in #1194
  • Due to changes in the Numpy C-API AdvancedIncSubtensor1 no longer has a C-implementation when paired with numpy>=2.0. This operator shows up in the gradient graph of advanced indexing operations, which could now be slower when using the C backend.
• Raise explicitly on Python methods that are incompatible with lazy variables by @ricardoV94 in #1190

New Features 🎉

• Speedup implementation of multivariate_normal and allow method of covariance decomposition by @ricardoV94 in #1203
• Compute pushforward (Rop) via double application of pullback (Lop) and fix Scan and Max gradient bugs by @ricardoV94 in #1207
• Add LAPACK overloads for all variants of solve in Numba backend by @jessegrabowski in #1146
• Implement multinomial and inverse gamma in numba backend by @ricardoV94 in #1223
• Implement gradient for vector repetitions by @ricardoV94 in #1192

Bugfixes 🐛

• Rewrite reshapes that only expand or squeeze dims by @ricardoV94 in #1200
• Fix Blockwise and RandomVariable in Numba with repeated arguments by @ricardoV94 in #1222

Maintenance 🔧

• Simplify implementation of tile by @ricardoV94 in #1187
• Remove accidental print statements by @ricardoV94 in #1193
• PyTorch inline constants in dispatch to avoid graph breaks by @ricardoV94 in #1118
• Speedup scalar Op python implementations by @ricardoV94 in #1169
• Only do reshapes in tensordot when needed by @ricardoV94 in #1202

New Contributors

• @brendan-m-murphy made their first contribution in #1194

Full Changelog: rel-2.27.1...rel-2.28.0

rel-2.27.1

What's Changed

New Features 🎉

• Add rewrite for 1 ** x = 1 by @jessegrabowski in #1179
• Allow overwriting existing rewrites during registration by @AdvH039 in #1119
• Rewrite batched dots that do not reduce as multiplication by @ricardoV94 in #1178

Bugfixes 🐛

• Fix numba 0.61 compatibility by @ricardoV94 in #1186
• Fix einsum bug by @ricardoV94 in #1185

Documentation 📖

• Fix typo in installation instructions for conda-forge by @inclinedadarsh in #1173
• Small tweaks the prng doc by @ricardoV94 in #1183
• Fix broken docs references by @ricardoV94 in #1180

Maintenance 🔧

• Make BLAS flags check lazy and more actionable by @ricardoV94 in #1165

New Contributors

• @inclinedadarsh made their first contribution in #1173
• @AdvH039 made their first contribution in #1119

Full Changelog: rel-2.27.0...rel-2.27.1

rel-2.27.0

What's Changed

Major Changes 🛠

• Deprecate redundant utilities for extracting constants by @ricardoV94 in #1046

New Features 🎉

• Implement ScalarLoop in torch backend by @Ch0ronomato in #958
• Pytensor-native interpolation functions by @jessegrabowski in #1141

Bugfixes 🐛

• More robust check for multiple integer indices in numba ravel_multidimensional_idx rewrites by @ricardoV94 in #1139
• Fix dispatch of Elemwised ScalarLoop in Numba backend by @ricardoV94 in #1137
• Use more conservative fastmath flags in numba backend by @ricardoV94 in #1147
• Respect predefined modes in get_default_mode by @ricardoV94 in #1166

Documentation 📖

• Add example gallery to docs by @jessegrabowski in #1150

Maintenance 🔧

• Make OpFromGraph available at root scope by @ricardoV94 in #1113
• Speedup pytensor import by @ricardoV94 in #1170

Full Changelog: rel-2.26.4...rel-2.27.0

rel-2.26.4

What's Changed

New Features 🎉

• Allow for more elemwise torch functions using broadcast_tensor and vmap by @Ch0ronomato in #1032
• Torch dispatch for scipy-like functions and Softplus by @Ch0ronomato in #1066
• Support consecutive integer vector indexing in Numba backend by @ricardoV94 in #1106
• Support multidimensional boolean set/inc_subtensor in Numba via rewrite by @ricardoV94 in #1108
• Add Numba implementation of Blockwise by @ricardoV94 in #1015

Documentation 📖

• Cross link across random variable docs by @ricardoV94 in #1091

Maintenance 🔧

• Slogdet returns naive expression and is optimized later by @tanish1729 in #1041
• Do not require explicit RNG in verify_grad by @aibit0111 in #1093
• Track generated functions for torch compile by @Ch0ronomato in #1094
• Reduce jitted function overhead by @ricardoV94 in #1101
• Improve performance of CAReduce in Numba backend by @ricardoV94 in #1109

New Contributors

• @aibit0111 made their first contribution in #1093

Full Changelog: rel-2.26.3...rel-2.26.4

rel-2.26.3

What's Changed

New Features 🎉

• Implement Kve Op and Kv helper by @ricardoV94 in #1081

Maintenance 🔧

• Use old ld64 linker on MacOS >= 15.0 by @lucianopaz in #1083
• Support on_unused_input for string parameter names in eval by @tvwenger in #1085

Full Changelog: rel-2.26.2...rel-2.26.3

rel-2.26.2

What's Changed

Bugfixes 🐛

• Fix bug in local_div_switch_sink rewrite by @ricardoV94 in #1080

Full Changelog: rel-2.26.1...rel-2.26.2

rel-2.26.1

What's Changed

New Features 🎉

• Implement unconditional constant_folding rewrite by @ricardoV94 in #1068

Full Changelog: rel-2.26.0...rel-2.26.1