Skip to content

P
pytensor
提交 1ea24b64 authored 作者: Pascal Lamblin's avatar Pascal Lamblin
浏览文件
操作

Modifications for 0.5 release

上级 f1ef401b
隐藏空白字符变更
内嵌 并排
正在显示 包含 214 行增加 和 427 行删除
EMAIL.0.5.txt deleted 100644 → 0
浏览文件 @ f1ef401b
===========================
Announcing Theano 0.5
===========================
Upgrading to Theano 0.5 is recommended for everyone, but you should first make
sure that your code does not raise deprecation warnings with Theano 0.4.1.
Otherwise, in one case the results can change. In other cases, the warnings are
turned into errors (see below for details).
For those using the bleeding edge version in the
git repository, we encourage you to update to the `0.5` tag.
What's New
----------
Important changes:
* Moved to github: http://github.com/Theano/Theano/
* Old trac ticket moved to assembla ticket: http://www.assembla.com/spaces/theano/tickets
* Theano vision: http://deeplearning.net/software/theano/introduction.html#theano-vision (Many people)
* Theano with GPU works in some cases on Windows now. Still experimental. (Sebastian Urban)
* See the Interface changes.
Interface Behavior Change (was deprecated and generated a warning since Theano 0.3 released Nov. 23rd, 2010):
* The current default value of the parameter axis of
theano.{max,min,argmax,argmin,max_and_argmax} is now the same as
numpy: None. i.e. operate on all dimensions of the tensor. (Frédéric Bastien, Olivier Delalleau)
Interface Features Removed (most were deprecated):
* The string modes FAST_RUN_NOGC and STABILIZE are not accepted. They were accepted only by theano.function().
Use Mode(linker='c|py_nogc') or Mode(optimizer='stabilize') instead.
* tensor.grad(cost, wrt) now always returns an object of the "same type" as wrt
(list/tuple/TensorVariable). (Ian Goodfellow, Olivier)
* A few tag.shape and Join.vec_length left have been removed. (Frederic)
* The .value attribute of shared variables is removed, use shared.set_value()
or shared.get_value() instead. (Frederic)
* Theano config option "home" is not used anymore as it was redundant with "base_compiledir".
If you use it, Theano will now raise an error. (Olivier D.)
* scan interface changes: (Razvan Pascanu)
- The use of `return_steps` for specifying how many entries of the output
to return has been removed. Instead, apply a subtensor to the output
returned by scan to select a certain slice.
- The inner function (that scan receives) should return its outputs and
updates following this order:
[outputs], [updates], [condition].
One can skip any of the three if not used, but the order has to stay unchanged.
Interface bug fixes:
* Rop in some case should have returned a list of one Theano variable, but returned the variable itself. (Razvan)
New deprecation (will be removed in Theano 0.6, warning generated if you use them):
* tensor.shared() renamed to tensor._shared(). You probably want to call theano.shared() instead! (Olivier D.)
New features:
* Adding 1D advanced indexing support to inc_subtensor and set_subtensor (James Bergstra)
* tensor.{zeros,ones}_like now support the dtype param as numpy (Frederic)
* Added configuration flag "exception_verbosity" to control the verbosity of exceptions (Ian)
* theano-cache list: list the content of the theano cache (Frederic)
* theano-cache unlock: remove the Theano lock (Olivier)
* tensor.ceil_int_div to compute ceil(a / float(b)) (Frederic)
* MaxAndArgMax.grad now works with any axis (The op supports only 1 axis) (Frederic)
* used by tensor.{max,min,max_and_argmax}
* tensor.{all,any} (Razvan)
* tensor.roll as numpy: (Matthew Rocklin, David Warde-Farley)
* Theano with GPU works in some cases on Windows now. Still experimental. (Sebastian Urban)
* IfElse now allows to have a list/tuple as the result of the if/else branches.
* They must have the same length and corresponding type (Razvan)
* Argmax output dtype is now int64 instead of int32. (Olivier)
* Added the element-wise operation arccos. (Ian)
* Added sparse dot with dense grad output. (Yann Dauphin)
* Optimized to Usmm and UsmmCscDense in some case (Yann)
* Note: theano.dot and theano.sparse.structured_dot() always had a gradient with the same sparsity pattern as the inputs.
The new theano.sparse.dot() has a dense gradient for all inputs.
* GpuAdvancedSubtensor1 supports broadcasted dimensions. (Frederic)
New optimizations:
* AdvancedSubtensor1 reuses preallocated memory if available (scan, c|py_nogc linker) (Frederic)
* tensor_variable.size (as numpy) computes the product of the shape elements. (Olivier)
* sparse_variable.size (as scipy) computes the number of stored values. (Olivier)
* dot22, dot22scalar work with complex. (Frederic)
* Generate Gemv/Gemm more often. (James)
* Remove scan when all computations can be moved outside the loop. (Razvan)
* scan optimization done earlier. This allows other optimizations to be applied. (Frederic, Guillaume, Razvan)
* exp(x) * sigmoid(-x) is now correctly optimized to the more stable form sigmoid(x). (Olivier)
* Added Subtensor(Rebroadcast(x)) => Rebroadcast(Subtensor(x)) optimization. (Guillaume)
* Made the optimization process faster. (James)
* Allow fusion of elemwise when the scalar op needs support code. (James)
* Better opt that lifts transpose around dot. (James)
Bug fixes (the result changed):
* On CPU, if the convolution had received explicit shape information, they where not checked at runtime.
This caused wrong result if the input shape was not the one expected. (Frederic, reported by Sander Dieleman)
* Scan grad when the input of scan has sequences of different lengths. (Razvan, reported by Michael Forbes)
* Scan.infer_shape now works correctly when working with a condition for the number of loops.
In the past, it returned n_steps as the length, which is not always true. (Razvan)
* Theoretical bug: in some case we could have GPUSum return bad value.
We were not able to reproduce this problem
* patterns affected ({0,1}*nb dim, 0 no reduction on this dim, 1 reduction on this dim):
01, 011, 0111, 010, 10, 001, 0011, 0101 (Frederic)
* div by zero in verify_grad. This hid a bug in the grad of Images2Neibs. (James)
* theano.sandbox.neighbors.Images2Neibs grad was returning a wrong value.
The grad is now disabled and returns an error. (Frederic)
Crashes fixed:
* T.mean crash at graph building time. (Ian)
* "Interactive debugger" crash fix. (Ian, Frederic)
* Do not call gemm with strides 0, some blas refuse it. (Pascal Lamblin)
* Optimization crash with gemm and complex. (Frederic)
* GPU crash with elemwise. (Frederic)
* Compilation crash with amdlibm and the GPU. (Frederic)
* IfElse crash. (Frederic)
* Execution crash fix in AdvancedSubtensor1 on 32 bit computers. (Pascal)
* GPU compilation crash on MacOS X. (Olivier)
* Support for OSX Enthought Python Distribution 7.x. (Graham Taylor, Olivier)
* When the subtensor inputs had 0 dimensions and the outputs 0 dimensions. (Frederic)
* Crash when the step to subtensor was not 1 in conjunction with some optimization. (Frederic, reported by Olivier Chapelle)
* fix dot22scalar cast of integer scalars (Justin Bayer, Frédéric, Olivier)
Known bugs:
* CAReduce with nan in inputs don't return the good output (`Ticket <https://www.assembla.com/spaces/theano/tickets/763>`_).
* This is used in tensor.{max,mean,prod,sum} and in the grad of PermuteRowElements.
* If you do grad of grad of scan you can have wrong results in some cases.
Sandbox:
* cvm interface more consistent with current linker. (James)
* vm linker has a callback parameter. (James)
* review/finish/doc: diag/extract_diag. (Arnaud Bergeron, Frederic, Olivier)
* review/finish/doc: AllocDiag/diag. (Arnaud, Frederic, Guillaume)
* review/finish/doc: MatrixInverse, matrix_inverse. (Razvan)
* review/finish/doc: matrix_dot. (Razvan)
* review/finish/doc: det (determinent) op. (Philippe Hamel)
* review/finish/doc: Cholesky determinent op. (David)
* review/finish/doc: ensure_sorted_indices. (Li Yao)
* review/finish/doc: spectral_radius_boud. (Xavier Glorot)
* review/finish/doc: sparse sum. (Valentin Bisson)
Sandbox New features (not enabled by default):
* CURAND_RandomStreams for uniform and normal (not picklable, GPU only) (James)
Documentation:
* Many updates. (Many people)
* Updates to install doc on MacOS. (Olivier)
* Updates to install doc on Windows. (David, Olivier)
* Added how to use scan to loop with a condition as the number of iteration. (Razvan)
* Added how to wrap in Theano an existing python function (in numpy, scipy, ...). (Frederic)
* Refactored GPU installation of Theano. (Olivier)
Others:
* Better error messages in many places. (David, Ian, Frederic, Olivier)
* PEP8 fixes. (Many people)
* New min_informative_str() function to print graph. (Ian)
* Fix catching of exception. (Sometimes we catched interupt) (Frederic, David, Ian, Olivier)
* Better support for uft string. (David)
* Fix pydotprint with a function compiled with a ProfileMode (Frederic)
* Was broken with change to the profiler.
* Warning when people have old cache entries. (Olivier)
* More tests for join on the GPU and CPU. (Frederic)
* Don't request to load the GPU module by default in scan module. (Razvan)
* Fixed some import problems.
* Filtering update. (James)
* The buidbot now raises optimization errors instead of just printing a warning. (Frederic)
* On Windows, the default compiledir changed to be local to the computer/user and not transferred with roaming profile. (Sebastian Urban)
Reviewers (alphabetical order):
* David, Frederic, Ian, James, Olivier, Razvan
This is a major release, with lots of new features, bug fixes, and some
interface changes (deprecated or potentially misleading features were
removed). The upgrade is recommended for everybody, unless you rely on
deprecated features that have been removed.
Download
--------
You can download Theano from http://pypi.python.org/pypi/Theano.
Description
-----------
Theano is a Python library that allows you to define, optimize, and
efficiently evaluate mathematical expressions involving
multi-dimensional arrays. It is built on top of NumPy. Theano
features:
* tight integration with NumPy: a similar interface to NumPy's.
numpy.ndarrays are also used internally in Theano-compiled functions.
* transparent use of a GPU: perform data-intensive computations up to
140x faster than on a CPU (support for float32 only).
* efficient symbolic differentiation: Theano can compute derivatives
for functions of one or many inputs.
* speed and stability optimizations: avoid nasty bugs when computing
expressions such as log(1+ exp(x)) for large values of x.
* dynamic C code generation: evaluate expressions faster.
* extensive unit-testing and self-verification: includes tools for
detecting and diagnosing bugs and/or potential problems.
Theano has been powering large-scale computationally intensive
scientific research since 2007, but it is also approachable
enough to be used in the classroom (IFT6266 at the University of Montreal).
Resources
---------
About Theano:
http://deeplearning.net/software/theano/
Theano-related projects:
http://github.com/Theano/Theano/wiki/Related-projects
About NumPy:
http://numpy.scipy.org/
About SciPy:
http://www.scipy.org/
Machine Learning Tutorial with Theano on Deep Architectures:
http://deeplearning.net/tutorial/
Acknowledgments
---------------
I would like to thank all contributors of Theano. For this particular
release, people names have been added next to what they did.
Also, thank you to all NumPy and Scipy developers as Theano builds on
their strengths.
All questions/comments are always welcome on the Theano
mailing-lists ( http://deeplearning.net/software/theano/#community )
EMAIL.txt
浏览文件 @ 1ea24b64
===========================
Announcing Theano 0.5rc2
Announcing Theano 0.5
===========================
## You can select and adapt one of the following templates.
## Basic text for major version release:
This is a release for a major version, with lots of new
features, bug fixes, and some interface changes (deprecated or
potentially misleading features were removed).
Upgrading to Theano 0.5 is recommended for everyone, but you should first make
sure that your code does not raise deprecation warnings with Theano 0.4.1.
Otherwise, in one case the results can change. In other cases, the warnings are
turned into errors (see below for details).
For those using the bleeding edge version in the
git repository, we encourage you to update to the `0.5` tag.
## Basic text for major version release candidate:
This is a release candidate for a major version, with lots of new
features, bug fixes, and some interface changes (deprecated or
potentially misleading features were removed).
The upgrade is recommended for developpers who want to help test and
report bugs, or want to use new features now. If you have updated
to 0.5rc1, you are highly encouraged to update to 0.5rc2. There are
more bug fixes and speed uptimization! But there is also a small new
interface change about sum of [u]int* dtype. Otherwise, users should
wait for the 0.5 release.
to 0.5rc1, you are highly encouraged to update to 0.5rc2.
For those using the bleeding edge version in the
git repository, we encourage you to update to the `0.5rc2` tag.
## Basic text for minor version release:
TODO
## Basic text for minor version release candidate:
TODO
What's New
----------
......@@ -83,16 +108,16 @@ Acknowledgments
I would like to thank all contributors of Theano. For this particular
release, many people have helped, notably (in alphabetical order):
Frédéric Bastien, Justin Bayer, Arnaud Bergerond, James Bergstra,
Valentin Bisson, Josh Bleecher Snyder, Yann Dauphin, Olivier Delalleau,
Guillaume Desjardins, Sander Dieleman, Xavier Glorot, Ian Goodfellow,
Philippe Hamel, Pascal Lamblin, Eric Laufer, Razvan Pascanu, Matthew
Rocklin, Graham Taylor, Sebastian Urban, David Warde-Farley, and Yao Li.
I would also like to thank users who submitted bug reports, notably
(this list is incomplete, please let us know if someone should be
added): Nicolas Boulanger-Lewandowski, Olivier Chapelle, Michael
Forbes, and Timothy Lillicrap.
Hani Almousli, Frédéric Bastien, Justin Bayer, Arnaud Bergeron, James
Bergstra, Valentin Bisson, Josh Bleecher Snyder, Yann Dauphin, Olivier
Delalleau, Guillaume Desjardins, Sander Dieleman, Xavier Glorot, Ian
Goodfellow, Philippe Hamel, Pascal Lamblin, Eric Laufer, Grégoire
Mesnil, Razvan Pascanu, Matthew Rocklin, Graham Taylor, Sebastian Urban,
David Warde-Farley, and Yao Li.
I would also like to thank users who submitted bug reports, notably:
Nicolas Boulanger-Lewandowski, Olivier Chapelle, Michael Forbes, Timothy
Lillicrap, and John Salvatier.
Also, thank you to all NumPy and Scipy developers as Theano builds on
their strengths.
......
NEWS.txt
浏览文件 @ 1ea24b64
.. _NEWS:
Since 0.5rc2
Bug fixes (the result changed):
* Fix a bug with Gemv and Ger on CPU, when used on vectors with negative
strides. Data was read from incorrect (and possibly uninitialized)
memory space. This bug was probably introduced in 0.5rc1. (Pascal L.)
Crashes fixes:
* More cases supported in AdvancedIncSubtensor1. (Olivier D.)
* Fix crash when a broadcasted constant was used as input of an
elemwise Op and needed to be upcasted to match the op's output.
(Reported by John Salvatier, fixed by Pascal L.)
Interface change:
* The Theano flag "nvcc.flags" is now included in the hard part of the key.
This mean that now we recompile all modules for each value of "nvcc.flags".
A change in "nvcc.flags" used to be ignored for module that were already
compiled. (Frederic B.)
* When using a GPU, detect faulty nvidia drivers. This was detected
when running Theano tests. Now this is always tested. Faulty
drivers results in in wrong results for reduce operations. (Frederic B.)
New features:
* Many infer_shape implemented on sparse matrices op. (David W.F.)
* Added theano.sparse.verify_grad_sparse to easily allow testing grad of
sparse op. It support testing the full and structured gradient.
* The keys in our cache now store the hash of constants and not the constant values
themselves. This is significantly more efficient for big constant arrays. (Frederic B.)
* 'theano-cache list' lists key files bigger than 1M (Frederic B.)
* 'theano-cache list' prints an histogram of the number of keys per compiled module (Frederic B.)
* 'theano-cache list' prints the number of compiled modules per op class (Frederic B.)
* The Theano flag "nvcc.fastmath" is now also used for the cuda_ndarray.cu file.
* Add the header_dirs to the hard part of the compilation key. This is
currently used only by cuda, but if we use library that are only headers,
this can be useful. (Frederic B.)
* Fixed a memory leak with shared variable (we kept a pointer to the original value) (Ian G.)
* Alloc, GpuAlloc are not always pre-computed (constant_folding optimization)
at compile time if all their inputs are constant.
(Frederic B., Pascal L., reported by Sander Dieleman)
* New Op tensor.sort(), wrapping numpy.sort (Hani Almousli)
=============
Release Notes
=============
If you have updated to 0.5rc1, you are highly encouraged to update to
0.5rc2. There are more bug fixes and speed uptimization! But there is
also a small new interface change about sum of [u]int* dtype.
Modifications in the trunk since the 0.4.1 release (August 12th, 2011)
======================================================================
Upgrading to Theano 0.5rc2 is recommended for everyone, but you should first make
sure that your code does not raise deprecation warnings with Theano 0.4.1.
Otherwise, in one case the results can change. In other cases, the warnings are
turned into errors (see below for details).
Theano 0.5 (23 February 2012)
=============================
Highlight:
* Moved to github: http://github.com/Theano/Theano/
* Old trac ticket moved to assembla ticket: http://www.assembla.com/spaces/theano/tickets
* Theano vision: http://deeplearning.net/software/theano/introduction.html#theano-vision (Many people)
* Theano with GPU works in some cases on Windows now. Still experimental. (Sebastian Urban)
* Faster dot() call: New/Better direct call to cpu and gpu ger, gemv, gemm and dot(vector, vector). (James, Frédéric, Pascal)
* Faster dot() call: New/Better direct call to cpu and gpu ger, gemv, gemm
and dot(vector, vector). (James, Frédéric, Pascal)
* C implementation of Alloc. (James, Pascal)
* theano.grad() now also work with sparse variable. (Arnaud)
* Macro to implement the Jacobian/Hessian with theano.tensor.{jacobian,hessian} (Razvan)
* See the Interface changes.
Interface Behavior Change (was deprecated and generated a warning since Theano 0.3 released Nov. 23rd, 2010):
Interface Behavior Changes:
* The current default value of the parameter axis of
theano.{max,min,argmax,argmin,max_and_argmax} is now the same as
numpy: None. i.e. operate on all dimensions of the tensor. (Frédéric Bastien, Olivier Delalleau)
numpy: None. i.e. operate on all dimensions of the tensor.
(Frédéric Bastien, Olivier Delalleau) (was deprecated and generated
a warning since Theano 0.3 released Nov. 23rd, 2010)
* The current output dtype of sum with input dtype [u]int* is now always [u]int64.
You can specify the output dtype with a new dtype parameter to sum.
The output dtype is the one using for the summation.
......@@ -82,10 +33,14 @@ Interface Behavior Change (was deprecated and generated a warning since Theano 0
The consequence is that the sum is done in a dtype with more precision than before.
So the sum could be slower, but will be more resistent to overflow.
This new behavior is the same as numpy. (Olivier, Pascal)
* When using a GPU, detect faulty nvidia drivers. This was detected
when running Theano tests. Now this is always tested. Faulty
drivers results in in wrong results for reduce operations. (Frederic B.)
Interface Features Removed (most were deprecated):
* The string modes FAST_RUN_NOGC and STABILIZE are not accepted. They were accepted only by theano.function().
* The string modes FAST_RUN_NOGC and STABILIZE are not accepted. They
were accepted only by theano.function().
Use Mode(linker='c|py_nogc') or Mode(optimizer='stabilize') instead.
* tensor.grad(cost, wrt) now always returns an object of the "same type" as wrt
(list/tuple/TensorVariable). (Ian Goodfellow, Olivier)
......@@ -103,13 +58,42 @@ Interface Features Removed (most were deprecated):
[outputs], [updates], [condition].
One can skip any of the three if not used, but the order has to stay unchanged.
Interface bug fixes:
* Rop in some case should have returned a list of one Theano variable, but returned the variable itself. (Razvan)
Interface bug fix:
* Rop in some case should have returned a list of one Theano variable,
but returned the variable itself. (Razvan)
New deprecation (will be removed in Theano 0.6, warning generated if you use them):
* tensor.shared() renamed to tensor._shared(). You probably want to call theano.shared() instead! (Olivier D.)
* tensor.shared() renamed to tensor._shared(). You probably want to
call theano.shared() instead! (Olivier D.)
Scan fix:
Bug fixes (incorrect results):
* On CPU, if the convolution had received explicit shape information,
they where not checked at runtime. This caused wrong result if the
input shape was not the one expected. (Frederic, reported by Sander
Dieleman)
* Theoretical bug: in some case we could have GPUSum return bad value.
We were not able to reproduce this problem
* patterns affected ({0,1}*nb dim, 0 no reduction on this dim, 1 reduction on this dim):
01, 011, 0111, 010, 10, 001, 0011, 0101 (Frederic)
* div by zero in verify_grad. This hid a bug in the grad of Images2Neibs. (James)
* theano.sandbox.neighbors.Images2Neibs grad was returning a wrong value.
The grad is now disabled and returns an error. (Frederic)
* An expression of the form "1 / (exp(x) +- constant)" was systematically matched to "1 / (exp(x) + 1)"
and turned into a sigmoid regardless of the value of the constant. A warning will be issued if your
code was affected by this bug. (Olivier, reported by Sander Dieleman)
* When indexing into a subtensor of negative stride (for instance, x[a:b:-1][c]),
an optimization replacing it with a direct indexing (x[d]) used an incorrect formula,
leading to incorrect results. (Pascal, reported by Razvan)
* The tile() function is now stricter in what it accepts to allow for better
error-checking/avoiding nonsensical situations. The gradient has been
disabled for the time being as it only implemented (incorrectly) one special
case. The `reps` argument must be a constant (not a tensor variable), and
must have the same length as the number of dimensions in the `x` argument;
this is now checked. (David)
Scan fixes:
* computing grad of a function of grad of scan (reported by Justin Bayer, fix by Razvan)
before : most of the time crash, but could be wrong value with bad number of dimensions (so a visible bug)
now : do the right thing.
......@@ -174,6 +158,26 @@ New features:
* tensor.tensordot can now be moved to GPU (Sander Dieleman,
Pascal, based on code from Tijmen Tieleman's gnumpy,
http://www.cs.toronto.edu/~tijmen/gnumpy.html)
* Many infer_shape implemented on sparse matrices op. (David W.F.)
* Added theano.sparse.verify_grad_sparse to easily allow testing grad of
sparse op. It support testing the full and structured gradient.
* The keys in our cache now store the hash of constants and not the constant values
themselves. This is significantly more efficient for big constant arrays. (Frederic B.)
* 'theano-cache list' lists key files bigger than 1M (Frederic B.)
* 'theano-cache list' prints an histogram of the number of keys per compiled module (Frederic B.)
* 'theano-cache list' prints the number of compiled modules per op class (Frederic B.)
* The Theano flag "nvcc.fastmath" is now also used for the cuda_ndarray.cu file.
* Add the header_dirs to the hard part of the compilation key. This is
currently used only by cuda, but if we use library that are only headers,
this can be useful. (Frederic B.)
* The Theano flag "nvcc.flags" is now included in the hard part of the key.
This mean that now we recompile all modules for each value of "nvcc.flags".
A change in "nvcc.flags" used to be ignored for module that were already
compiled. (Frederic B.)
* Alloc, GpuAlloc are not always pre-computed (constant_folding optimization)
at compile time if all their inputs are constant.
(Frederic B., Pascal L., reported by Sander Dieleman)
* New Op tensor.sort(), wrapping numpy.sort (Hani Almousli)
New optimizations:
......@@ -189,30 +193,6 @@ New optimizations:
* Better opt that lifts transpose around dot. (James)
Bug fixes (the result changed):
* On CPU, if the convolution had received explicit shape information, they where not checked at runtime.
This caused wrong result if the input shape was not the one expected. (Frederic, reported by Sander Dieleman)
* Theoretical bug: in some case we could have GPUSum return bad value.
We were not able to reproduce this problem
* patterns affected ({0,1}*nb dim, 0 no reduction on this dim, 1 reduction on this dim):
01, 011, 0111, 010, 10, 001, 0011, 0101 (Frederic)
* div by zero in verify_grad. This hid a bug in the grad of Images2Neibs. (James)
* theano.sandbox.neighbors.Images2Neibs grad was returning a wrong value.
The grad is now disabled and returns an error. (Frederic)
* An expression of the form "1 / (exp(x) +- constant)" was systematically matched to "1 / (exp(x) + 1)"
and turned into a sigmoid regardless of the value of the constant. A warning will be issued if your
code was affected by this bug. (Olivier, reported by Sander Dieleman)
* When indexing into a subtensor of negative stride (for instance, x[a:b:-1][c]),
an optimization replacing it with a direct indexing (x[d]) used an incorrect formula,
leading to incorrect results. (Pascal, reported by Razvan)
* The tile() function is now stricter in what it accepts to allow for better
error-checking/avoiding nonsensical situations. The gradient has been
disabled for the time being as it only implemented (incorrectly) one special
case. The `reps` argument must be a constant (not a tensor variable), and
must have the same length as the number of dimensions in the `x` argument;
this is now checked. (David)
Crashes fixed:
* T.mean crash at graph building time. (Ian)
* "Interactive debugger" crash fix. (Ian, Frederic)
......@@ -238,6 +218,11 @@ Crashes fixed:
when matrices had non-unit stride in both dimensions (CPU and GPU),
or when matrices had negative strides (GPU only). In those cases,
we are now making copies. (Pascal)
* More cases supported in AdvancedIncSubtensor1. (Olivier D.)
* Fix crash when a broadcasted constant was used as input of an
elemwise Op and needed to be upcasted to match the op's output.
(Reported by John Salvatier, fixed by Pascal L.)
* Fixed a memory leak with shared variable (we kept a pointer to the original value) (Ian G.)
Known bugs:
......
doc/NEWS.txt
浏览文件 @ 1ea24b64
.. _NEWS:
Since 0.5rc2
* Fixed a memory leak with shared variable (we kept a pointer to the original value)
* Alloc, GpuAlloc are not always pre-computed (constant_folding optimization) at compile time if all their inputs are constant
* The keys in our cache now store the hash of constants and not the constant values themselves. This is significantly more efficient for big constant arrays.
* 'theano-cache list' lists key files bigger than 1M
* 'theano-cache list' prints an histogram of the number of keys per compiled module
* 'theano-cache list' prints the number of compiled modules per op class
=============
Release Notes
=============
If you have updated to 0.5rc1, you are highly encouraged to update to
0.5rc2. There are more bug fixes and speed uptimization! But there is
also a small new interface change about sum of [u]int* dtype.
Modifications in the trunk since the 0.4.1 release (August 12th, 2011)
======================================================================
Upgrading to Theano 0.5rc2 is recommended for everyone, but you should first make
sure that your code does not raise deprecation warnings with Theano 0.4.1.
Otherwise, in one case the results can change. In other cases, the warnings are
turned into errors (see below for details).
Theano 0.5 (23 February 2012)
=============================
Highlight:
* Moved to github: http://github.com/Theano/Theano/
* Old trac ticket moved to assembla ticket: http://www.assembla.com/spaces/theano/tickets
* Theano vision: http://deeplearning.net/software/theano/introduction.html#theano-vision (Many people)
* Theano with GPU works in some cases on Windows now. Still experimental. (Sebastian Urban)
* Faster dot() call: New/Better direct call to cpu and gpu ger, gemv, gemm and dot(vector, vector). (James, Frédéric, Pascal)
* Faster dot() call: New/Better direct call to cpu and gpu ger, gemv, gemm
and dot(vector, vector). (James, Frédéric, Pascal)
* C implementation of Alloc. (James, Pascal)
* theano.grad() now also work with sparse variable. (Arnaud)
* Macro to implement the Jacobian/Hessian with theano.tensor.{jacobian,hessian} (Razvan)
* See the Interface changes.
Interface Behavior Change (was deprecated and generated a warning since Theano 0.3 released Nov. 23rd, 2010):
Interface Behavior Changes:
* The current default value of the parameter axis of
theano.{max,min,argmax,argmin,max_and_argmax} is now the same as
numpy: None. i.e. operate on all dimensions of the tensor. (Frédéric Bastien, Olivier Delalleau)
numpy: None. i.e. operate on all dimensions of the tensor.
(Frédéric Bastien, Olivier Delalleau) (was deprecated and generated
a warning since Theano 0.3 released Nov. 23rd, 2010)
* The current output dtype of sum with input dtype [u]int* is now always [u]int64.
You can specify the output dtype with a new dtype parameter to sum.
The output dtype is the one using for the summation.
There is no warning in previous Theano version about this.
The consequence is that the sum is done in a dtype with more precession then before.
The consequence is that the sum is done in a dtype with more precision than before.
So the sum could be slower, but will be more resistent to overflow.
This new behavior is the same as numpy. (Olivier, Pascal)
* When using a GPU, detect faulty nvidia drivers. This was detected
when running Theano tests. Now this is always tested. Faulty
drivers results in in wrong results for reduce operations. (Frederic B.)
Interface Features Removed (most were deprecated):
* The string modes FAST_RUN_NOGC and STABILIZE are not accepted. They were accepted only by theano.function().
* The string modes FAST_RUN_NOGC and STABILIZE are not accepted. They
were accepted only by theano.function().
Use Mode(linker='c|py_nogc') or Mode(optimizer='stabilize') instead.
* tensor.grad(cost, wrt) now always returns an object of the "same type" as wrt
(list/tuple/TensorVariable). (Ian Goodfellow, Olivier)
......@@ -71,22 +58,51 @@ Interface Features Removed (most were deprecated):
[outputs], [updates], [condition].
One can skip any of the three if not used, but the order has to stay unchanged.
Interface bug fixes:
* Rop in some case should have returned a list of one Theano variable, but returned the variable itself. (Razvan)
Interface bug fix:
* Rop in some case should have returned a list of one Theano variable,
but returned the variable itself. (Razvan)
New deprecation (will be removed in Theano 0.6, warning generated if you use them):
* tensor.shared() renamed to tensor._shared(). You probably want to call theano.shared() instead! (Olivier D.)
* tensor.shared() renamed to tensor._shared(). You probably want to
call theano.shared() instead! (Olivier D.)
Scan fix:
* computing grad of a function of grad of scan(reported by ?, Razvan)
before : most of the time crash, but could be wrong value with bad number of dimensions(so a visible bug)
Bug fixes (incorrect results):
* On CPU, if the convolution had received explicit shape information,
they where not checked at runtime. This caused wrong result if the
input shape was not the one expected. (Frederic, reported by Sander
Dieleman)
* Theoretical bug: in some case we could have GPUSum return bad value.
We were not able to reproduce this problem
* patterns affected ({0,1}*nb dim, 0 no reduction on this dim, 1 reduction on this dim):
01, 011, 0111, 010, 10, 001, 0011, 0101 (Frederic)
* div by zero in verify_grad. This hid a bug in the grad of Images2Neibs. (James)
* theano.sandbox.neighbors.Images2Neibs grad was returning a wrong value.
The grad is now disabled and returns an error. (Frederic)
* An expression of the form "1 / (exp(x) +- constant)" was systematically matched to "1 / (exp(x) + 1)"
and turned into a sigmoid regardless of the value of the constant. A warning will be issued if your
code was affected by this bug. (Olivier, reported by Sander Dieleman)
* When indexing into a subtensor of negative stride (for instance, x[a:b:-1][c]),
an optimization replacing it with a direct indexing (x[d]) used an incorrect formula,
leading to incorrect results. (Pascal, reported by Razvan)
* The tile() function is now stricter in what it accepts to allow for better
error-checking/avoiding nonsensical situations. The gradient has been
disabled for the time being as it only implemented (incorrectly) one special
case. The `reps` argument must be a constant (not a tensor variable), and
must have the same length as the number of dimensions in the `x` argument;
this is now checked. (David)
Scan fixes:
* computing grad of a function of grad of scan (reported by Justin Bayer, fix by Razvan)
before : most of the time crash, but could be wrong value with bad number of dimensions (so a visible bug)
now : do the right thing.
* gradient with respect to outputs using multiple taps(reported by Timothy, fix by Razvan)
* gradient with respect to outputs using multiple taps (reported by Timothy, fix by Razvan)
before : it used to return wrong values
now : do the right thing.
Note: The reported case of this bug was happening in conjunction with the
save optimization of scan that give run time errors. So if you didn't
manually disable the same memory optimization(number in the list4),
manually disable the same memory optimization (number in the list4),
you are fine if you didn't manually request multiple taps.
* Rop of gradient of scan (reported by Timothy and Justin Bayer, fix by Razvan)
before : compilation error when computing R-op
......@@ -97,7 +113,7 @@ Scan fix:
* Scan grad when the input of scan has sequences of different lengths. (Razvan, reported by Michael Forbes)
* Scan.infer_shape now works correctly when working with a condition for the number of loops.
In the past, it returned n_steps as the length, which is not always true. (Razvan)
* Scan.infer_shape crash fix. (Reported by ?, Razvan)
* Scan.infer_shape crash fix. (Razvan)
New features:
* AdvancedIncSubtensor grad defined and tested (Justin Bayer)
......@@ -128,20 +144,40 @@ New features:
* We also support the "theano_version" substitution.
* IntDiv c code (faster and allow this elemwise to be fused with other elemwise) (Pascal)
* Internal filter_variable mechanism in Type. (Pascal, Ian)
* Ifelse work on sparse.
* Make use of gpu shared variable more transparent with theano.function updates and givens parameter.
* Ifelse works on sparse.
* It makes use of gpu shared variable more transparent with theano.function updates and givens parameter.
* Added a_tensor.transpose(axes) axes is optional (James)
* theano.tensor.transpose(a_tensor, kwargs) We where ignoring kwargs, now it is used as the axes.
* a_CudaNdarray_object[*] = int, now work (Frederic)
* a_CudaNdarray_object[*] = int, now works (Frederic)
* tensor_variable.size (as numpy) computes the product of the shape elements. (Olivier)
* sparse_variable.size (as scipy) computes the number of stored values. (Olivier)
* sparse_variable[N, N] now works (Li Yao, Frederic)
* sparse_variable[M:N, O:P] now works (Li Yao, Frederic)
* Warning: M, N, O, and P should be Python int or scalar tensor variables,
in particular, None is not well-supported.
* sparse_variable[M:N, O:P] now works (Li Yao, Frederic, Pascal)
M, N, O, and P can be Python int or scalar tensor variables, None, or
omitted (sparse_variable[:, :M] or sparse_variable[:M, N:] work).
* tensor.tensordot can now be moved to GPU (Sander Dieleman,
Pascal, based on code from Tijmen Tieleman's gnumpy,
http://www.cs.toronto.edu/~tijmen/gnumpy.html)
* Many infer_shape implemented on sparse matrices op. (David W.F.)
* Added theano.sparse.verify_grad_sparse to easily allow testing grad of
sparse op. It support testing the full and structured gradient.
* The keys in our cache now store the hash of constants and not the constant values
themselves. This is significantly more efficient for big constant arrays. (Frederic B.)
* 'theano-cache list' lists key files bigger than 1M (Frederic B.)
* 'theano-cache list' prints an histogram of the number of keys per compiled module (Frederic B.)
* 'theano-cache list' prints the number of compiled modules per op class (Frederic B.)
* The Theano flag "nvcc.fastmath" is now also used for the cuda_ndarray.cu file.
* Add the header_dirs to the hard part of the compilation key. This is
currently used only by cuda, but if we use library that are only headers,
this can be useful. (Frederic B.)
* The Theano flag "nvcc.flags" is now included in the hard part of the key.
This mean that now we recompile all modules for each value of "nvcc.flags".
A change in "nvcc.flags" used to be ignored for module that were already
compiled. (Frederic B.)
* Alloc, GpuAlloc are not always pre-computed (constant_folding optimization)
at compile time if all their inputs are constant.
(Frederic B., Pascal L., reported by Sander Dieleman)
* New Op tensor.sort(), wrapping numpy.sort (Hani Almousli)
New optimizations:
......@@ -157,36 +193,12 @@ New optimizations:
* Better opt that lifts transpose around dot. (James)
Bug fixes (the result changed):
* On CPU, if the convolution had received explicit shape information, they where not checked at runtime.
This caused wrong result if the input shape was not the one expected. (Frederic, reported by Sander Dieleman)
* Theoretical bug: in some case we could have GPUSum return bad value.
We were not able to reproduce this problem
* patterns affected ({0,1}*nb dim, 0 no reduction on this dim, 1 reduction on this dim):
01, 011, 0111, 010, 10, 001, 0011, 0101 (Frederic)
* div by zero in verify_grad. This hid a bug in the grad of Images2Neibs. (James)
* theano.sandbox.neighbors.Images2Neibs grad was returning a wrong value.
The grad is now disabled and returns an error. (Frederic)
* An expression of the form "1 / (exp(x) +- constant)" was systematically matched to "1 / (exp(x) + 1)"
and turned into a sigmoid regardless of the value of the constant. A warning will be issued if your
code was affected by this bug. (Olivier, reported by Sander Dieleman)
* When indexing into a subtensor of negative stride (for instance, x[a:b:-1][c]),
an optimization replacing it with a direct indexing (x[d]) used an incorrect formula,
leading to incorrect results. (Pascal, reported by Razvan)
* The tile() function is now stricter in what it accepts to allow for better
error-checking/avoiding nonsensical situations. The gradient has been
disabled for the time being as it only implemented (incorrectly) one special
case. The `reps` argument must be a constant (not a tensor variable), and
must have the same length as the number of dimensions in the `x` argument;
this is now checked. (David)
Crashes fixed:
* T.mean crash at graph building time. (Ian)
* "Interactive debugger" crash fix. (Ian, Frederic)
* Do not call gemm with strides 0, some blas refuse it. (Pascal Lamblin)
* Optimization crash with gemm and complex. (Frederic)
* GPU crash with elemwise. (Frederic)
* GPU crash with elemwise. (Frederic, some reported by Chris Currivan)
* Compilation crash with amdlibm and the GPU. (Frederic)
* IfElse crash. (Frederic)
* Execution crash fix in AdvancedSubtensor1 on 32 bit computers. (Pascal)
......@@ -199,7 +211,18 @@ Crashes fixed:
* Fix runtime crash in gemm, dot22. FB
* Fix on 32bits computer: make sure all shape are int64.(Olivier)
* Fix to deque on python 2.4 (Olivier)
* Fix crash when not using c code(or using DebugMode)(not used by default) with numpy 1.6*. Numpy have a bug in the reduction code that make it crash. ufunc.reduce (Pascal)
* Fix crash when not using c code (or using DebugMode) (not used by
default) with numpy 1.6*. Numpy has a bug in the reduction code that
made it crash. (Pascal)
* Crashes of blas functions (Gemv on CPU; Ger, Gemv and Gemm on GPU)
when matrices had non-unit stride in both dimensions (CPU and GPU),
or when matrices had negative strides (GPU only). In those cases,
we are now making copies. (Pascal)
* More cases supported in AdvancedIncSubtensor1. (Olivier D.)
* Fix crash when a broadcasted constant was used as input of an
elemwise Op and needed to be upcasted to match the op's output.
(Reported by John Salvatier, fixed by Pascal L.)
* Fixed a memory leak with shared variable (we kept a pointer to the original value) (Ian G.)
Known bugs:
......@@ -242,26 +265,30 @@ Documentation:
Others:
* Better error messages in many places. (Many people)
* PEP8 fixes. (Many people)
* Add a warning about numpy bug with subtensor with more then 2**32 elemenent(TODO, more explicit)
* Added Scalar.ndim=0 and ScalarSharedVariable.ndim=0 (simplify code)(Razvan)
* Add a warning about numpy bug when using advanced indexing on a
tensor with more than 2**32 elements (the resulting array is not
correctly filled and ends with zeros). (Pascal, reported by David WF)
* Added Scalar.ndim=0 and ScalarSharedVariable.ndim=0 (simplify code) (Razvan)
* New min_informative_str() function to print graph. (Ian)
* Fix catching of exception. (Sometimes we used to catch interrupts) (Frederic, David, Ian, Olivier)
* Better support for uft string. (David)
* Better support for utf string. (David)
* Fix pydotprint with a function compiled with a ProfileMode (Frederic)
* Was broken with change to the profiler.
* Warning when people have old cache entries. (Olivier)
* More tests for join on the GPU and CPU. (Frederic)
* Don't request to load the GPU module by default in scan module. (Razvan)
* Do not request to load the GPU module by default in scan module. (Razvan)
* Fixed some import problems. (Frederic and others)
* Filtering update. (James)
* On Windows, the default compiledir changed to be local to the computer/user and not transferred with roaming profile. (Sebastian Urban)
* On Windows, the default compiledir changed to be local to the
computer/user and not transferred with roaming profile. (Sebastian
Urban)
* New theano flag "on_shape_error". Defaults to "warn" (same as previous behavior):
it prints a warning when an error occurs when inferring the shape of some apply node.
The other accepted value is "raise" to raise an error when this happens. (Frederic)
* The buidbot now raises optimization/shape errors instead of just printing a warning. (Frederic)
* better pycuda tests (Frederic)
* check_blas.py now accept the shape and the number of iteration as parameter (Frederic)
* Fix opt warning when the opt ShapeOpt is disabled(enabled by default) (Frederic)
* Fix opt warning when the opt ShapeOpt is disabled (enabled by default) (Frederic)
* More internal verification on what each op.infer_shape return. (Frederic, James)
* Argmax dtype to int64 (Olivier)
* Improved docstring and basic tests for the Tile Op (David).
......
doc/conf.py
浏览文件 @ 1ea24b64
......@@ -53,7 +53,7 @@ copyright = '2008--2012, LISA lab'
# The short X.Y version.
version = '0.5'
# The full version, including alpha/beta/rc tags.
release = '0.5rc2'
release = '0.5'
# There are two options for replacing |today|: either, you set today to some
# non-false value, then it is used:
......
setup.py
浏览文件 @ 1ea24b64
......@@ -48,7 +48,7 @@ PLATFORMS = ["Windows", "Linux", "Solaris", "Mac OS-X", "Unix"]
MAJOR = 0
MINOR = 5
MICRO = 0
SUFFIX = "rc2" # Should be blank except for rc's, betas, etc.
SUFFIX = "" # Should be blank except for rc's, betas, etc.
ISRELEASED = False
VERSION = '%d.%d.%d%s' % (MAJOR, MINOR, MICRO, SUFFIX)
......
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论