提交 c735f1d5 authored 作者: notoraptor's avatar notoraptor

Improve NEWS list.

Simplify warn.ignore_bug_before parameters. Rewrite "cudnn" to "cuDNN".
上级 8e1ad95a
...@@ -16,18 +16,17 @@ Highlight: ...@@ -16,18 +16,17 @@ Highlight:
- More graph optimization (faster execution and smaller graph, so more readable) - More graph optimization (faster execution and smaller graph, so more readable)
- Less c code compilation - Less c code compilation
- Better Python 3.5 support - Better Python 3.5 support
- Better numpy 1.12.0rc2 support - Better numpy 1.12 support
- Support newer Mac and Windows version - Support newer Mac and Windows version
- Conda packages for Mac, Linux and Windows - Conda packages for Mac, Linux and Windows
- Theano scripts now works on Windows - Theano scripts now works on Windows
- scan with checkpoint (trade off between speed and memory usage, useful for long sequences) - scan with checkpoint (trade off between speed and memory usage, useful for long sequences)
- Added a bool dtype - Added a bool dtype
- Added a deprecation message for the old backend.
- New back-end: - New back-end:
- float16 storage - float16 storage
- print pci bus id - better mapping between theano device number and nvidia-smi number, using the PCI bus ID of graphic cards
- More pooling support on GPU when cudnn isn't there. - More pooling support on GPU when cuDNN isn't there.
- ignore_border=False is now implemented for pooling. - ignore_border=False is now implemented for pooling.
...@@ -50,14 +49,14 @@ Convolution updates: ...@@ -50,14 +49,14 @@ Convolution updates:
- Dilated convolution - Dilated convolution
GPU: GPU:
- CuDNN: support versoin 5.1 and wrap batch normalization (2d and 3d) and RNN functions - cuDNN: support versoin 5.1 and wrap batch normalization (2d and 3d) and RNN functions
- Multiple-GPU, synchrone update (via platoon, use NCCL) - Multiple-GPU, synchrone update (via platoon, use NCCL)
- GpuAdvancedSubtensor in new back-end - GpuAdvancedSubtensor in new back-end
- Gemv(matrix-vector product) speed up for special shape - Gemv(matrix-vector product) speed up for special shape
- Support for MaxAndArgMax for some axis combination - Support for MaxAndArgMax for some axis combination
- Support for solve (using cusolver), erfinv and erfcinv - Support for solve (using cusolver), erfinv and erfcinv
- cublas gemv workaround when we reduce on an axis with a dimensions size of 0 - cublas gemv workaround when we reduce on an axis with a dimensions size of 0
- Warn user that some cudnn algorithms may produce unexpected results in certain environments - Warn user that some cuDNN algorithms may produce unexpected results in certain environments
for convolution backward filter operations. for convolution backward filter operations.
New features: New features:
......
...@@ -23,18 +23,17 @@ Highlight: ...@@ -23,18 +23,17 @@ Highlight:
- More graph optimization (faster execution and smaller graph, so more readable) - More graph optimization (faster execution and smaller graph, so more readable)
- Less c code compilation - Less c code compilation
- Better Python 3.5 support - Better Python 3.5 support
- Better numpy 1.12.0rc2 support - Better numpy 1.12 support
- Support newer Mac and Windows version - Support newer Mac and Windows version
- Conda packages for Mac, Linux and Windows - Conda packages for Mac, Linux and Windows
- Theano scripts now works on Windows - Theano scripts now works on Windows
- scan with checkpoint (trade off between speed and memory usage, useful for long sequences) - scan with checkpoint (trade off between speed and memory usage, useful for long sequences)
- Added a bool dtype - Added a bool dtype
- Added a deprecation message for the old backend.
- New back-end: - New back-end:
- float16 storage - float16 storage
- print pci bus id - better mapping between theano device number and nvidia-smi number, using the PCI bus ID of graphic cards
- More pooling support on GPU when cudnn isn't there. - More pooling support on GPU when cuDNN isn't there.
- ignore_border=False is now implemented for pooling. - ignore_border=False is now implemented for pooling.
Interface changes: Interface changes:
...@@ -54,14 +53,14 @@ Convolution updates: ...@@ -54,14 +53,14 @@ Convolution updates:
GPU: GPU:
- CuDNN: support versoin 5.1 and wrap batch normalization (2d and 3d) and RNN functions - cuDNN: support versoin 5.1 and wrap batch normalization (2d and 3d) and RNN functions
- Multiple-GPU, synchrone update (via platoon, use NCCL) - Multiple-GPU, synchrone update (via platoon, use NCCL)
- GpuAdvancedSubtensor in new back-end - GpuAdvancedSubtensor in new back-end
- Gemv(matrix-vector product) speed up for special shape - Gemv(matrix-vector product) speed up for special shape
- Support for MaxAndArgMax for some axis combination - Support for MaxAndArgMax for some axis combination
- Support for solve (using cusolver), erfinv and erfcinv - Support for solve (using cusolver), erfinv and erfcinv
- cublas gemv workaround when we reduce on an axis with a dimensions size of 0 - cublas gemv workaround when we reduce on an axis with a dimensions size of 0
- Warn user that some cudnn algorithms may produce unexpected results in certain environments - Warn user that some cuDNN algorithms may produce unexpected results in certain environments
for convolution backward filter operations. for convolution backward filter operations.
New features: New features:
......
Markdown 格式
0%
您添加了 0 到此讨论。请谨慎行事。
请先完成此评论的编辑!
注册 或者 后发表评论