projects / platform / core / ml / nntrainer.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 63cac02) | patch
[Bug] Fix generating nan values in tensor
authorDonghyeon Jeong <djeong20@illinois.edu>
Tue, 1 Aug 2023 02:42:00 +0000 (11:42 +0900)
committerJijoong Moon <jijoong.moon@samsung.com>
Mon, 21 Aug 2023 06:29:23 +0000 (15:29 +0900)
commit08f74d0040e19026bd66164647657cdfebb81285
treef3b6b2d2c52a5d7053ad2c14b7fb21734a26dfb1tree | snapshot
parent63cac02f37f0558fb1497265cee86cd378fac628commit | diff
[Bug] Fix generating nan values in tensor
- Gradient tensor values are inconsistently set to NaN
- NaN values caused incorrect backwarding in Neural Net
- Replacing malloc with calloc prevents memory allocation with value set to NaN

Signed-off-by: Donghyeon Jeong <djeong20@illinois.edu>
nntrainer/tensor/memory_pool.cpp diff | blob | history
Domain: Machine Learning / ML Framework;