Prepare for 0.12.0 Release #8263
Conversation
NEWS.md
Outdated
| @@ -1,5 +1,30 @@ | |||
| MXNet Change Log | |||
| ================ | |||
| ## 0.12.0 | |||
| ### - New Features - Sparse Tensor Support | |||
There was a problem hiding this comment.
Don't need the leading -
NEWS.md
Outdated
| - `NDArray` and `Symbol` now supports "fluent" methods. You can now use `x.exp()` etc instead of `mx.nd.exp(x)` or `mx.sym.exp(x)` | ||
| - Added `mx.rtc.CudaModule` for writing and running CUDA kernels from python | ||
| - Added `multi_precision` option to optimizer for easier float16 training | ||
| ### - Performance |
There was a problem hiding this comment.
Why are two other performance improvements removed?
There was a problem hiding this comment.
Added 1 back as discussed - " Enabled JIT compilation. Autograd and Gluon hybridize now use less memory and has faster speed. Performance is almost the same with old symbolic style code"
NEWS.md
Outdated
| - Added `mx.autograd.grad` and experimental second order gradient support (though most operators don't support second order gradient yet) | ||
| - Added `ConvLSTM` etc to `gluon.contrib` | ||
| - Autograd now supports cross-device graphs. Use `x.copyto(mx.gpu(i))` and `x.copyto(mx.cpu())` to do computation on multiple devices. | ||
| ### - Other New Features |
There was a problem hiding this comment.
Also list limited support for fancy indexing. x[idx_arr0, idx_arr1, ..., idx_arrn] is now supported. Full support coming soon in next release. Checkout master to get a preview
| @@ -21,7 +21,7 @@ | |||
| # Built files are stored in $built | |||
| # Version numbers are stored in $tag_list. | |||
| # Version numbers are ordered from latest to old and final one is master. | |||
| tag_list="0.11.0.rc3 master" | |||
| tag_list="0.12.0 0.11.0 master" | |||
There was a problem hiding this comment.
@sandeep-krishnamurthy @kevinthesun
Should this change be made now or after the RC is final?
NEWS.md
Outdated
| ## 0.12.0 | ||
| ### New Features - Sparse Tensor Support | ||
| - Added comprehensive support for sparse matrices. See help on `mx.sym.sparse` and `mx.nd.sparse` for more info. | ||
| - Limited support for fancy indexing. x[idx_arr0, idx_arr1, ..., idx_arrn] is now supported. Full support coming soon in next release. Checkout master to get a preview. |
There was a problem hiding this comment.
fancy indexing is for normal NDArray
There was a problem hiding this comment.
@eric-haibin-lin Can you please edit the line and let me know how you want to rephrase it?I will update the PR
There was a problem hiding this comment.
I am just saying that this doesn't belong to sparse tensor
There was a problem hiding this comment.
Moved to other features. Done.
NEWS.md
Outdated
| ### Performance | ||
| - Enabled JIT compilation. Autograd and Gluon hybridize now use less memory and has faster speed. Performance is almost the same with old symbolic style code. | ||
| - Full support for NVidia Volta GPU Architecture and Cuda 9. Training is up to 3.5x faster than Pascal when using float16. | ||
| ### - API Changes |
There was a problem hiding this comment.
why is there a dash before API Changes?
NEWS.md
Outdated
| @@ -1,11 +1,38 @@ | |||
| MXNet Change Log | |||
| ================ | |||
| ## 0.12.0 | |||
| ### New Features - Sparse Tensor Support | |||
| - Added comprehensive support for sparse matrices. See help on `mx.sym.sparse` and `mx.nd.sparse` for more info. | |||
There was a problem hiding this comment.
- Added limited cpu support for two sparse formats for
SymbolandNDArray-CSRNDArrayandRowSparseNDArray - Added a sparse dot product operator and many element-wise sparse operators
- Added a data iterator for sparse data input -
LibSVMIter - Added three optimizers for sparse gradient updates:
Ftrl,SGDandAdam - Added
pushandrow_sparse_pullwithRowSparseNDArrayin distributed kvstore
There was a problem hiding this comment.
Should I replace the existing point or add subpoints?
There was a problem hiding this comment.
Yeah please replace the existing point. comprehensive support is misleading since we dont have gpu support. Thanks!
There was a problem hiding this comment.
Ah, all the formats are gone.. :(
NEWS.md
Outdated
| - Added a data iterator for sparse data input - LibSVMIter | ||
| - Added three optimizers for sparse gradient updates: Ftrl, SGD and Adam | ||
| - Added push and row_sparse_pull with RowSparseNDArray in distributed kvstore | ||
| - For more information see [full release notes](https://cwiki.apache.org/confluence/display/MXNET/MXNet+0.12.0+Release+Notes) |
There was a problem hiding this comment.
-
- For more information see full release notes
Why is this put under sparse? Should we move it to the end?
- For more information see full release notes
* Preparing for 0.12.0 Release * Correcting Syntax in NEWS.md * Adding to NEWS.md * Adding to NEWS.md * build_all_version.sh should be updated only after RC passes * Edits * Changes to NEWS.md * formatting
* Preparing for 0.12.0 Release * Correcting Syntax in NEWS.md * Adding to NEWS.md * Adding to NEWS.md * build_all_version.sh should be updated only after RC passes * Edits * Changes to NEWS.md * formatting
* Preparing for 0.12.0 Release * Correcting Syntax in NEWS.md * Adding to NEWS.md * Adding to NEWS.md * build_all_version.sh should be updated only after RC passes * Edits * Changes to NEWS.md * formatting
Description
Checklist
Essentials
make lint)Changes
Comments
Coming up in In another PR before tagging the RC -
Updated README.md
Updated NEWS.md if needed