提交 · e6d07bcb4b261312768abba2dd7cdbc8a2de21dc · testgroup / pytensor

07 4月, 2022 10 次提交
- Delete disabled gpu-related config options · e6d07bcb
  由 Brandon T. Willard 提交于 3月 28, 2022
  
  e6d07bcb
- Remove fast_compile_gpu tag · 640c12f3
  由 Brandon T. Willard 提交于 3月 28, 2022
  
  640c12f3
- Move TensorType import to top level · 5415f12d
  由 Maxim Kochurov 提交于 3月 14, 2022
  
  5415f12d
- Remove unnecessary gpuarray comment from InplaceElemwiseOptimizer · 664c8a0d
  由 Brandon T. Willard 提交于 3月 28, 2022
  
  664c8a0d
- Remove PoolingMode_t · 5eb34a3c
  由 Brandon T. Willard 提交于 3月 28, 2022
  
  5eb34a3c
- Remove GPU options from DeviceParam · 3b0c4946
  由 Maxim Kochurov 提交于 3月 14, 2022
  
  3b0c4946
- Remove gpuarray documentation · 32b79040
  由 Maxim Kochurov 提交于 3月 14, 2022
  
  32b79040
- Remove gpuarray dependencies throughout the codebase · 0e3182d1
  由 Maxim Kochurov 提交于 3月 14, 2022
  
  0e3182d1
- Remove aesara.gpuarray · 2a5fc594
  由 Maxim Kochurov 提交于 3月 14, 2022
  
  2a5fc594
- Remove tests.gpuarray · b3ce3640
  由 Maxim Kochurov 提交于 3月 11, 2022
  
  b3ce3640
06 4月, 2022 2 次提交
- Load the sparse shared constructor support at package-level · c803c67e
  由 Brandon T. Willard 提交于 4月 04, 2022
  
  c803c67e
- Update pre-commit version of black · b4442683
  由 Brandon T. Willard 提交于 4月 04, 2022
  
  b4442683
01 4月, 2022 1 次提交
- Add JAX dispatch for SolveTriangular OP, resolves #846 · e60cfb57
  由 Rob Zinkov 提交于 3月 31, 2022
  
  e60cfb57
24 3月, 2022 2 次提交

Handle negative axis in `squeeze` and do not ignore invalid axis · ef256e52

由提交于 2月 24, 2022

* Previous implementation would silently ignore explicit axis for dimensions that were not broadcastable.
* Added note about difference in behavior between Aesara and Numpy functions

Closes #830
Co-authored-by: Brandon T. Willard <971601+brandonwillard@users.noreply.github.com>

ef256e52

Validate axis in `CAReduce.make_node` · 8f96d930
由 Ricardo 提交于 2月 24, 2022
```
Closes #677
```
8f96d930

23 3月, 2022 1 次提交
- Remove old warning when using a dict for updates · 39a455d3
  由 Ricardo 提交于 3月 22, 2022
  
  39a455d3
22 3月, 2022 1 次提交
- Use type name in Elemwise and CAReduce __str__ implementations · c8922630
  由 Brandon T. Willard 提交于 3月 20, 2022
  
  c8922630
20 3月, 2022 3 次提交
- Fix an issue with non-homogeneous shape tuples in Numba's BroadcastTo · 9023e2b3
  由 Brandon T. Willard 提交于 3月 18, 2022
  
  9023e2b3
- Add Numba implementation for DirichletRV · 081967d3
  由 Brandon T. Willard 提交于 3月 18, 2022
  
  081967d3
- Rename function and remove unnecessary param for Numba CategoricalRV · 87751407
  由 Brandon T. Willard 提交于 3月 18, 2022
  
  87751407
19 3月, 2022 5 次提交
- Fix typing issues in aesara.scalar.math · 27423150
  由 Brandon T. Willard 提交于 3月 18, 2022
  
  27423150
- Fix typing issues in aesara.scalar.basic · ed62da3e
  由 Brandon T. Willard 提交于 3月 18, 2022
  
  ed62da3e
- Rename Scalar to ScalarType · 470b9d60
  由 Brandon T. Willard 提交于 3月 18, 2022
  
  470b9d60
- Add HasDataType mixin to Scalar · 9136eb66
  由 Brandon T. Willard 提交于 3月 18, 2022
  
  9136eb66
- Added linspace, geomspace and logspace to aesara.tensor.extra_ops · 1d3a1623
  由 kc611 提交于 3月 17, 2022
  
  1d3a1623
18 3月, 2022 4 次提交
- Use params in Numba's default Op.perform implementation · 980c4c2c
  由 Brandon T. Willard 提交于 3月 18, 2022
  
  980c4c2c
- Add node comments preceding lines produced by aesara.link.utils.fgraph_to_python · 9a1ab3e8
  由 Brandon T. Willard 提交于 3月 17, 2022
  
  9a1ab3e8
- Convert SciPy scalar function inputs to acceptable dtypes in Numba implementations · 4ccf184f
  由 Brandon T. Willard 提交于 3月 17, 2022
  
  4ccf184f
- Fix Numba CAReduce infinite identity values for non-floats · 033bf332
  由 Brandon T. Willard 提交于 3月 17, 2022
  
  033bf332
17 3月, 2022 4 次提交
- Convert shapes to scalar tuples in Numba BroadcastTo implementation · d0cb1374
  由 Brandon T. Willard 提交于 3月 16, 2022
  
  d0cb1374
- Use tuple_setitem from numba_basic to simplify mocking · a6b2b0cb
  由 Brandon T. Willard 提交于 3月 16, 2022
  
  a6b2b0cb
- Remove unnecessary Subtensors from broadcast_to · cb4da6ed
  由 Brandon T. Willard 提交于 3月 16, 2022
  
  cb4da6ed
- Use new JAX index update approach when available · 605b609f
  由 Brandon T. Willard 提交于 3月 16, 2022
  
  605b609f
16 3月, 2022 3 次提交
- Rename SparseType to SparseTensorType · 94f5ddfd
  由 Brandon T. Willard 提交于 3月 15, 2022
  
  94f5ddfd
- Warn when dense conversion is used by sparse tensor methods · b8c1c463
  由 Anatoly 提交于 1月 28, 2022
  
  b8c1c463
- Make sparse tensor types extend the existing tensor types · fdf2a23a
  由 Brandon T. Willard 提交于 2月 14, 2021
  
  fdf2a23a
14 3月, 2022 4 次提交
- Fix `DiffOp` output type when input has partially known shape · 31ab8fdd
  由 Ricardo 提交于 3月 11, 2022
  
  31ab8fdd
- Fix `DiffOp` `infer_shape` when `n` is larger than axis dimension · fde62163
  由 Ricardo 提交于 3月 11, 2022
```
Other changes:
* Test negative axis
* Add test for faulty gradient behavior when n is larger than input size
* Add test for NotImplementedError for gradients of ndim != 1
* Remove vague aesara.function call in tests
```
  fde62163
- Add rewrites to remove unnecessary expm1 operations · 40b51621
  由 Ricardo 提交于 1月 17, 2022
  
  40b51621
- Remove unnecessary type checks in `local_exp_log*` rewrites · e8c2782f
  由 Ricardo 提交于 1月 18, 2022
  
  e8c2782f