https://github.com/cudamat/cudamat

sort by:
Revision Author Date Message Commit Date
21baa0a Merge pull request #101 from cloudmercato/master Fix missing optimization level in nvcc call 28 May 2020, 16:28:22 UTC
130a1a0 Fix missing optimization level in nvcc call 28 May 2020, 16:18:57 UTC
6f4dbc7 Merge pull request #75 from f0k/improve-sum-performance Improve performance of dot() and sum() 09 September 2016, 18:10:51 UTC
d157cfa Cheaper CUDAMatrix.ones assignment, without host copy 08 September 2016, 16:17:13 UTC
f8e7b3e Use gemv if first dot() argument is a vector 08 September 2016, 16:17:13 UTC
a7ea270 Use gemv if second dot() argument is a vector 08 September 2016, 15:31:38 UTC
de44205 Merge pull request #73 from AminHP/filter_function Added correlate function 23 August 2016, 10:20:06 UTC
931ef92 Add correlate() function 18 August 2016, 13:56:02 UTC
06778f7 Merge pull request #70 from BCSharp/python3_last_cuda_error Making get_last_cuda_error() Python3-compliant 29 June 2016, 11:00:15 UTC
494aae6 get_last_cuda_error() Python3-compliant 28 June 2016, 18:47:18 UTC
e28a463 Merge pull request #68 from BCSharp/init_random_unicode Issue #67: Python 3 support in init_random 28 June 2016, 16:01:09 UTC
f2065b3 More comprehensive error checks when loading rnd_multipliers_32bit.txt 28 June 2016, 04:23:34 UTC
0141b1e Issue #67: File open error checking in init_random 23 June 2016, 17:41:51 UTC
3c4d1d7 Fix for Issue #67: Proper handling of unicode paths in Python 3 23 June 2016, 17:30:27 UTC
d7fd4eb Mention Python 3 compatibility in INSTALL file 21 March 2016, 22:48:44 UTC
59405ff Merge pull request #42 from f0k/win-support Windows support 04 August 2015, 14:57:06 UTC
6565e63 Update setup.py to work with MSVC 04 August 2015, 12:04:37 UTC
bf6a8a8 Merge pull request #52 from cudamat/change-test-directory Install instructions: run test and benchmark from different directory 29 July 2015, 09:05:04 UTC
135518a Install instructions: run test and benchmark from different directory 29 July 2015, 09:02:13 UTC
5d6a74f Support import under Windows 28 July 2015, 12:41:02 UTC
a12a1ff Support compilation under Windows 28 July 2015, 12:33:27 UTC
af7d9ca Python 3 compatibility Closes #43 16 January 2015, 13:31:46 UTC
8667218 Merge pull request #41 from NervanaSystems/osx_anaconda_linking Strip arch flags when linking on OSX. Closes #39. 12 January 2015, 10:26:12 UTC
78f961f Strip arch flags when linking on OSX. See #39 08 January 2015, 19:03:42 UTC
3eee7cf Merge pull request #32 from f0k/update-readme Updated installation instructions, added contribution instructions 03 December 2014, 10:28:17 UTC
b57b4f6 Updated installation instructions, added contribution instructions 02 December 2014, 15:51:45 UTC
20efac2 Merge pull request #31 from NervanaSystems/pip_package make cudamat pip installable. 02 December 2014, 10:06:39 UTC
0a061d9 Merge pull request #1 from f0k/more-setuptools Use nvcc for linking 01 December 2014, 19:59:10 UTC
aafb1ba Allow NVCCFLAGS to customize compiler invocations similar to CFLAGS, CXXFLAGS etc. 01 December 2014, 11:34:30 UTC
4b78bbe Use nvcc for linking to avoid CUDA library lookup 01 December 2014, 11:25:44 UTC
5c4e1d0 continue when LD_LIBRARY_PATH not set. 29 November 2014, 03:16:30 UTC
27bc863 Create custom nvcc compiler extension. 29 November 2014, 03:06:18 UTC
d843b62 keep learn submodule functions separate. 13 November 2014, 18:17:02 UTC
15523cc python3 compatibility, doc updates. 09 November 2014, 02:56:42 UTC
a6a6025 make cudamat pip installable. 08 November 2014, 01:39:31 UTC
a346369 Merge pull request #30 from ebattenberg/max-ones Add "max_ones" argument to cublas_init + PEP8 cleanup 26 August 2014, 16:38:59 UTC
ad7f91a remove errant error check in test_cudamat.py - also cleanup spaces on blank lines (PEP8) 25 August 2014, 23:52:43 UTC
7c40aba PEP8: Reduce all lines to max 80 characters 24 August 2014, 21:50:42 UTC
a10d4f1 update exception raising syntax to new style Old format is deprecated in PEP8 24 August 2014, 21:38:20 UTC
469a4a0 remove unused variable 24 August 2014, 21:36:56 UTC
c243e6f PEP8: fix spacing - around keyword args - between top-level functions 24 August 2014, 21:35:56 UTC
a48240f cleanup unused imports 24 August 2014, 21:24:01 UTC
5103859 add vim support to .gitignore 24 August 2014, 21:18:22 UTC
4f41581 fix max_ones doc, remove blank line. 24 August 2014, 21:17:45 UTC
360ad96 remove tabs and remove spaces on blank lines 11 August 2014, 00:48:53 UTC
b16e170 document max_ones argument to cublas_init 11 August 2014, 00:45:38 UTC
0954dc9 add ones-reduction allocation checking 25 July 2014, 20:24:51 UTC
8e33077 add max_ones arg to cublas init 25 July 2014, 20:24:51 UTC
874622a add error check for init cublas 25 July 2014, 20:24:51 UTC
fffb8a9 Merge pull request #28 from f0k/test-scripts-run Make test_* scripts run directly 22 July 2014, 12:55:50 UTC
81c6937 Make test_* scripts run the tests when called directly instead of using the nosetests script 22 July 2014, 12:51:08 UTC
d613d60 Merge pull request #22 from untom/where Added cudamat.where function. 08 November 2013, 21:15:59 UTC
35c81c4 Improved documentation of cudamat.where 06 November 2013, 14:52:10 UTC
71bb712 Added cudamat.where 05 November 2013, 14:15:11 UTC
ab70937 Merge pull request #21 from f0k/update-readme2 Change git clone url in README from SSH to HTTPS 08 October 2013, 15:43:24 UTC
6a275cc Changed git clone url in README from SSH to HTTPS 08 October 2013, 15:41:28 UTC
c10ae4c Merge pull request #20 from f0k/update-readme Update README with content from google code project page 08 October 2013, 15:23:50 UTC
18744d1 Updated README file with content from google code project page 08 October 2013, 15:14:09 UTC
82cb237 Merge pull request #15 from f0k/dynamic-gridsize Choose block count dynamically 30 September 2013, 15:09:34 UTC
93dadf5 Choose CUDA grid dimensions dynamically for element-wise operations 11 September 2013, 18:17:53 UTC
b026c7a Merge pull request #17 from f0k/benchmark Add benchmark script timing different operations 11 September 2013, 17:38:18 UTC
ad8c6cd Merge pull request #16 from f0k/fix-reciprocal-warning Fix warning about double/float conversion in kReciprocal() kernel 11 September 2013, 17:36:58 UTC
76afda0 Merge pull request #6 from f0k/slice-host Allow CUDAMatrix.slice() to produce a view into the CPU copy as well 11 September 2013, 17:36:18 UTC
9d0d5ca Merge pull request #10 from f0k/div-by-row Added div_by_col() and div_by_row() 11 September 2013, 17:35:41 UTC
a4f1f83 Fix warning about double/float conversion in kReciprocal() kernel 11 September 2013, 16:49:59 UTC
45171d4 Add benchmark script timing different operations 11 September 2013, 16:47:42 UTC
6b2138a Resolve upstream conflicts hindering automatic merge 10 September 2013, 17:35:54 UTC
4b48827 Fixed whitespace problem hindering automatic merge 10 September 2013, 17:19:07 UTC
a1ee4c1 Merge pull request #13 from f0k/fix-so-path Fix inferred path for .so files if importing in same directory 09 September 2013, 20:48:20 UTC
170a554 Merge pull request #8 from f0k/reshape-more Make CUDAMatrix.reshape() consistently reshape all versions of the matrix 09 September 2013, 20:44:00 UTC
5651945 Merge pull request #11 from f0k/fix-euclid-norm Fix euclid_norm() error handling when matrix is not on device 09 September 2013, 20:29:14 UTC
d1be3e8 Merge pull request #12 from f0k/fix-equals-sync Fix missing SYNC_THREADS check in equals() 09 September 2013, 20:28:17 UTC
b14ced1 Merge pull request #14 from f0k/min-max-more Add a couple of minimum/maximum related functions 09 September 2013, 20:26:29 UTC
1ad1ff0 Merge pull request #4 from f0k/leverage-sgemm Expose alpha and beta arguments of cublasSgemm 09 September 2013, 20:21:01 UTC
a1a8b51 Merge pull request #9 from f0k/soft-threshold Add CUDAMatrix.apply_soft_threshold 09 September 2013, 20:19:31 UTC
bc88db9 Merge pull request #7 from f0k/matrix-copy provide CUDAMatrix.copy() 09 September 2013, 20:09:56 UTC
f11d722 Merge pull request #5 from f0k/init-nocopy Allow CUDAMatrix.__init__() to not copy the CPU array 09 September 2013, 20:08:34 UTC
abc60bd Added argmin() and argmax() 04 September 2013, 12:07:47 UTC
e6672f3 Added min() 04 September 2013, 11:04:16 UTC
9d4119e Fix inferred path for .so files if importing modules from python shell in same directory 04 September 2013, 10:55:03 UTC
3e9a813 Add axis=1 support to max() 04 September 2013, 10:52:32 UTC
d670b51 Added minimum() and maximum() 03 September 2013, 16:28:40 UTC
be45653 fixed missing SYNC_THREADS check in equals() 03 September 2013, 15:57:57 UTC
9188a82 Fixed euclid_norm() error handling when matrix is not on device 03 September 2013, 15:52:41 UTC
18c745e Added div_by_col() and div_by_row() 03 September 2013, 15:47:52 UTC
e33c499 Adds CUDAMatrix.apply_soft_threshold 26 August 2013, 16:41:14 UTC
f571ce7 Ensure CUDAMatrix.reshape() consistently reshapes all versions of the matrix (GPU, GPU transposed, CPU) 23 August 2013, 16:59:31 UTC
639a826 CUDAMatrix.__init__() can now be told to not copy the ndarray it is initialized from 23 August 2013, 16:50:22 UTC
b7bd4d5 provide CUDAMatrix.copy() 23 August 2013, 16:33:05 UTC
d43c519 CUDAMatrix.slice() can now produce a view into the CPU copy as well 23 August 2013, 16:30:48 UTC
87f01dd Save a multiplication for the momentum update in the NN example 23 August 2013, 14:12:50 UTC
a671685 Exposed alpha and beta arguments of cublasSgemm where possible; added mean() 23 August 2013, 13:56:38 UTC
0e225ce Merge pull request #3 from f0k/manhattan-norm manhattan_norm() and allfinite() 23 August 2013, 13:45:53 UTC
8ef1b26 Added CUDAMatrix.manhattan_norm() and CUDAMatrix.allfinite() 23 August 2013, 13:12:07 UTC
f8ad9ea Merge pull request #1 from f0k/absolute-so-path Load cudamat libraries by absolute path on Linux/Mac 23 August 2013, 04:51:51 UTC
17d0b30 Load cudamat libraries by absolute path on Linux/Mac, making it easier to install 22 August 2013, 17:23:10 UTC
3803b3b Initial commit. 21 August 2013, 19:55:35 UTC
d1f9a58 Initial commit 21 August 2013, 19:27:47 UTC
back to top