* Fix ConstructSparseFromList.infer_shape, (Pascal L., reported by Rami Al-Rfou')
* (introduced in the development version after 0.6rc3 release) (Frederic B.)
Reduction that upcast the input on no axis (ex: call theano.sum() on a scalar when the original dtype isn't float64 or [u]int64). It produce bad results as we don't upcast the inputs in the code, we just copy it.
* Fix some case of theano.clone() with we git it replacement of x that is a function of x. (Razvan P., reported by Akio Takano)
New Features:
* Python 3.3 compatible (abalkin, Gabe Schwartz, Frederic B.)
...
...
@@ -84,10 +86,15 @@ New Features:
* Add gpu max for pattern (0, 1) and added all gpu max pattern for gpu min. (Frederic B.)
* Add GpuEye (Frederic B.)
* Make GpuCrossentropySoftmaxArgmax1HotWithBias and GpuCrossentropySoftmax1HotWithBiasDx work for bigger inputs (Frederic B., reported by Ryan Price)
* Finish and move out of sandbox theano.sparse.basic.true_dot (Nicolas Bouchard, Frederic B.)
And document all sparse dot variant.
Interface Deprecation (a warning is printed):
* The mode ProfileMode is now deprecated, use the Theano flag profile=True to remplace it.
* New theano.sparse_grad() interface to get the sparse grad of a_tensor[an_int_vector]. (Frederic B.)
This can speed up the sparse computation when a small fraction of a_tensor is taken.
Deprecate the old interface for this. (Frederic B.)
Interface Change:
* Add -m32 or -m64 in the module cache key and add the python bitwidth in the compiledir path. (Pascal L.)
...
...
@@ -125,6 +132,7 @@ Crash fixes:
* Crash fix in Scan following interface change in 0.6rc2 (Razvan P.)
* Crash fix on Scan (Razvan P.)
* Fix crash in Scan gradient related to compute_test_value (Frederic B., reported by Bitton Tenessi)
* Fix a scan optimization warning/error depending of Theano flags (Frederic B.)
* Fixed crash for unimplemented elemwise gradient (Olivier D., reported by Michael McNeil Forbes)
* Fix crash in the elemwise python code for some big shape with power of 2. (Sina Honari, Pascal L.)
* Fix compile and import errors on Windows (bbudescu)
...
...
@@ -149,9 +157,10 @@ Others:
* Typo/pep8 (Olivier D., Frederic B.)
* Update doc (Ben McCann)
* Doc compatibility guide (abalkin)
* Doc the MPI and load op (Frederic B.)
* Add manual instruction for OpenBLAS on Ubuntu by (Jianri Li )
* Doc fixes (Yaroslav Halchenko)
* Better error message (Ian Goodfellow)
* Better/more error message (Frederic B., Pascal L., Ian Goodfellow)
* More doc (Frederic B.)
* Fix Error reporting with GpuConv (Frederic B., reported by Heng Luo and Nicolas Pinto)
* Update BLAS compilation doc on windows to use OpenBLAS (Olivier D.)
...
...
@@ -164,8 +173,7 @@ Others:
* Added BLAS benchmark result (Frederic B., Ben McCann)
* Fix code comment (Hannes Schulz)
* More stable tests (Frederic B.)
* Add utt.asset_allclose(a, b) to have better error messge. (Frederic B.)
* Better error message (Pascal L., Frederic B.)
* Add utt.asset_allclose(a, b) to have better error message. (Frederic B.)
* Better error message with compute_test_value (Frederic, reported by John Salvatier)
* Stochastic order behavior fix (Frederic B.)
* Simpler initial graph for subtensor infer shape (Olivier D.)
...
...
@@ -185,22 +193,6 @@ Todo for the final release:
* https://github.com/Theano/Theano/pull/1450
* More stable test (Frederic B.)
* https://github.com/Theano/Theano/pull/1441
* On error during execution of an Theano function, now always print the apply node, its inputs shape, strides and types. If Theano flags exception_verbosity=high, also print debugprint(op, stop_on_name=True, print_type=True)
* https://github.com/Theano/Theano/pull/1238
Refers to the bug reported by Akio Takano on the mailing list. The problem refers to the case when one tries to replace x with a function of x, f(x).
The fix implies first replacing x with some tmp_x and then tmp_x with f(x).
