[pull] master from clab:master #1
8000
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Fixing equation mistake in documentation
[WIP] simplify Parameters in python
Distinguishes between multiline strings and comments
* dynet-1298. Add expr ** 2 and expr ** expr(not elementwise) support in python layer. * dynet-1298. Update pow description in _dynet.pyx
* dynet-1296. curand support for multi GPU. * add curandeng control inside GPU device * clean up global curandeng. * fix macro
* Fix huber_distance bug and make expression readable
Correct a minor typo in the python dynet.cg_version() documentatoin
* dynet-1320. Add / support for cdiv. * comment typo fix
* Fix swig bindings - move dynet::detail to expr.h so swig bindings can use it - two small fixes (include nodes.h, drop default parameter) * swig example: set learning rate instead of updateEpoch
Otherwise, memory of project_scratch is leaking and gradient_norm_scratch use will cause problems.
Currently, there is no indices checking in gradient_l2_norm_dev and sometimes the code could request elements beyond the end of the vector (when k1 = params.size() or k2 = lookup_params.size()). To fix this, we introduce a simple bounds check. params.size() > k1 lookup_params.size() > k2
…1629) If an ExpressionVector is garbage collected, the C++ memory for the vector and its expressions will be freed in C++. If any Scala Expressions that were derived from it with apply(n) are still alive, their C++ counterparts will have disappeared out from under them. The Scala Expression, when used, will try to access its C++ counterpart and cause the C++ code to crash and take down the JVM with it.
…rface (#1576) Its subclasses are marked in dynet_swig.i as using shared pointers: ``` %shared_ptr(dynet::ParameterStorage) %shared_ptr(dynet::LookupParameterStorage) ``` SWIG warns that the base class "is not similarly marked as a smart pointer". The line below quiets the warning and also prevents crahes (because of double deletes?). ``` %shared_ptr(dynet::ParameterStorageBase) ``` An issue was filed previously and this will soon link to it.
…1631) Bumps [pillow](https://github.com/python-pillow/Pillow) from 6.2.0 to 7.1.0. - [Release notes](https://github.com/python-pillow/Pillow/releases) - [Changelog](https://github.com/python-pillow/Pillow/blob/master/CHANGES.rst) - [Commits](python-pillow/Pillow@6.2.0...7.1.0) Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
This flag causes issues on M1 Macs, as clang does not support this flag on this platform. Instead, use -march-apple-m1.
…1645) Bumps [pillow](https://github.com/python-pillow/Pillow) from 7.1.0 to 9.0.0. - [Release notes](https://github.com/python-pillow/Pillow/releases) - [Changelog](https://github.com/python-pillow/Pillow/blob/main/CHANGES.rst) - [Commits](python-pillow/Pillow@7.1.0...9.0.0) --- updated-dependencies: - dependency-name: pillow dependency-type: direct:production ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
…-process training on a machine (across the time), which happens a lot in a cluster environment. (#1633) terminate called after throwing an instance of 'boost::interprocess::interprocess_exception' what(): Permission denied What happened is: User A run a multi-process training. A queue file with name "dynet_mp_work_queue" is created with each multi-process training, and is often not deleted properly due to manual termination (like CTRL+C). Now User B tries to run a multi-process training on the same machine, and could not write/delete the queue file. So the above error is thrown. Fix: Include process id in the queue file name, so different trainings by different users do not step on each other's toes.
* added implicit parameter to expression casting * added option for device specification for inputs functions like ones, zeros. * Added GPU (and device specification) support in csharp wrapper * Updated documentation for GPU support * Updated markdown * added input functions by dimensions instead of value * added LeCunUniform and NoamOpt - c++,python,sharp, fixed bugs in dynetsharp, adjusted transformer example to work with opennmt model * added mini-batching for dynet sharp! Also, fixed bugs, added another example, updated to work with new optimizers * Fixed NoamOpt learning-rate curve, verified result * Reverted not dynetsharp changes * Removed added HeInitializer and NoamOpt * Updated vcxProj to NetCore * Fixed shuffle
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot]. Want to support this open source service? Please star it : )