nntrainer
nntrainer copied to clipboard
nnstreamer
NNtrainer is Software Framework for Training Neural Network Models on Devices.
This is branch out for Tizen 8.0 M2 Release. Signed-off-by: jijoong.moon
This patch is for inference mode for a swap device. It adds "memory_swap_mode" property, which can handle inference mode when using swap device. The weights are not modified while inferencing,...
This patch implements unittest for checkpoint. Signed-off-by: Jiho Chu
This patch is for checkpoint tensor load/unload. For forward step, used unchecked node's tensors are reclaimed after used. These reclaimed tensors are reloaded while backward step if it's needed. And...
This patch adds CheckPoint property to layer node. Default value is checkpointed, which can store the layer data for the backward step. laoded and unloaded state are for unchecked nodes....
temp pool is for temporal tensor which is used for checkpoint mechanism. It reduces peak memory consumption while training, it only store checkpointed activation for the backward phase. unchecked activations...
Unlike other layers, embedding layer uses specific form of Tensor called `IndexedSlices` and this data contains the specific indicies of the Tensor that we are interested in. Thus, during the...
This patch adds chrome tracing feature for trace module. chrome tracing is GUI profile framework usually work with browser, but many uses it as common profile UI tool. This patch...
The trainable property was inserted with fault value with a transfer learning option.
We already support operations such as multiply or transpose at the tensor level, but if we support them at the layer level like addition_layer, it will be much more convenient...
Metadata
Owner
Metadata
NNtrainer is Software Framework for Training Neural Network Models on Devices.