summary |
shortlog | log |
commit |
commitdiff |
tree
first ⋅ prev ⋅ next
Inki Dae [Tue, 16 Nov 2021 07:12:07 +0000 (16:12 +0900)]
add save model api
[Version] : 0.1.0-0
[Issue type] : new feature
This api saves trained model data to a given model file.
Change-Id: Icc79b2138d5c8beecaef684e4cd540ffabd1af30
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Inki Dae [Thu, 9 Sep 2021 03:08:58 +0000 (12:08 +0900)]
drop model_file_path parameter from CreateModel()
NNTrainer needs only config file path so drop it,
and consider model_conf_path.
Change-Id: Iffb04b81905716a64e5a922689f6e6ca6cbe501b
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Inki Dae [Mon, 30 Aug 2021 04:47:52 +0000 (13:47 +0900)]
add DestroyOptimizer function
Change-Id: Ibafc6afc9e51ee91516d0b380fd54c1d0ea14c48
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Inki Dae [Thu, 26 Aug 2021 05:55:18 +0000 (14:55 +0900)]
update nntrainer backend interface
This patch updates below things
- implement to update dataset through a given callbacks - one is to send
training data to NNTrainer, and other for verifying it.
- fix ml_train_model_compile function call by setting only valid property
options.
- consider ml_train_model_run function call without any options.
. Regarding this function, NNTrainer has default property optinos so
in case of no given option, we can use the default ones.
Change-Id: I0bc25a491403149641f41931f43965efddbf15a8
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Inki Dae [Thu, 26 Aug 2021 05:51:11 +0000 (14:51 +0900)]
add BuildRequires to capi-ml-inference
training engine nntrainer backend needs capi-ml-inference api
so add the BuildRequires for it.
Change-Id: I009468d0c223dabe772d0b10fe7c18fd30c16a80
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Inki Dae [Fri, 20 Aug 2021 08:39:02 +0000 (17:39 +0900)]
do not use ml_train_dataset_create_with_generator function
ml_train_dataset_create_with_generator function has been deprecated
according to https://github.com/nnstreamer/nntrainer/blob/main/api/capi/include/nntrainer.h#L418
Use ml_train_dataset_create function instead of it.
Change-Id: I233631220cab8f085ae2a9e8bb3d5b18cf7f8c0e
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Inki Dae [Fri, 20 Aug 2021 06:48:35 +0000 (15:48 +0900)]
consider one more strings for all property options
Change-Id: Ic89a9428665575bdee6ac91a9aad40cbdff06cd3
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Inki Dae [Tue, 17 Aug 2021 11:15:06 +0000 (20:15 +0900)]
convert more layer types
Change-Id: Id8ac25ec946923973fcd36c6890fb68af3464021
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Inki Dae [Tue, 17 Aug 2021 11:14:30 +0000 (20:14 +0900)]
consider one more strings for layer option
Change-Id: I7b24d59c1416630c406d51bbd981a02a5026537f
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Inki Dae [Wed, 11 Aug 2021 08:00:27 +0000 (17:00 +0900)]
Add initial code
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Tizen Infrastructure [Mon, 9 Aug 2021 12:02:31 +0000 (12:02 +0000)]
Initial empty repository
projects / platform / upstream / training-engine-nntrainer.git / log