platform/core/ml/nntrainer.git
3 years ago[tensor_filter] Fix tensorfilter fail
Jihoon Lee [Tue, 24 Aug 2021 07:31:52 +0000 (16:31 +0900)]
[tensor_filter] Fix tensorfilter fail

This patch fixes tensor filter fail due to two critical bugs

- readModel was broken
- tensor getSharedDataTensor was broken

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[layer_node] bugfix/validate init_context
hyeonseok lee [Wed, 25 Aug 2021 07:12:32 +0000 (16:12 +0900)]
[layer_node] bugfix/validate init_context

 - Validate init_context before using the init_context in print function
 - Reorder adding default layer object to appcontext based on LayerType enum
 - Correct typo

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years ago[Dataset] Change finalize signature
Jihoon Lee [Wed, 11 Aug 2021 05:40:14 +0000 (14:40 +0900)]
[Dataset] Change finalize signature

This patch changes finalize signature to contain user_data,
user_data will be needed to replicate the current worker(with diffrent
arguments) in case of
MT_safty is not provided by the producer. Please note that this does not
change any current behavior but the preparation for the future
enhancements.

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[Dataset] Rework func dataset to samplewise
Jihoon Lee [Tue, 10 Aug 2021 10:56:15 +0000 (19:56 +0900)]
[Dataset] Rework func dataset to samplewise

**Changes proposed in this PR:**
- Fix bugs in Tensor::Map
- Add shortcut for the Tensor::updateBatch
- Add func_data_producer with sample

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[unittest] backbone property name update
Parichay Kapoor [Mon, 23 Aug 2021 06:02:25 +0000 (15:02 +0900)]
[unittest] backbone property name update

Update backbone property name for the layers unittests.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[CAPI] Open centroid KNN
Jihoon Lee [Wed, 18 Aug 2021 08:35:34 +0000 (17:35 +0900)]
[CAPI] Open centroid KNN

**Changes proposed in this PR:**
- Move centroid KNN to layer, delete centroid knn from the Application

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Cc: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[memorypool] Introducing memory pool
Parichay Kapoor [Tue, 10 Aug 2021 12:59:40 +0000 (21:59 +0900)]
[memorypool] Introducing memory pool

This patch introduces memory pool for tensors. All the tensors requiring
memory can ask memory pool for their memory. The memory pool collects
all the requirements, uses the provided memory planner to plan a memory
layout, allocates the required memory, and then finally returns the
allocated memory to the requesters.

There are certain limitations to this memory pool:
1. All the memory must be requested before layout planning and
allocation.
2. There is no support for freeing selectively allocated tensors
3. In order to request support for more tensors and their memories,
all previously allocated memory has to be freed.
4. MemoryPool for now allocates all its required memory in a single
block. Support for breaking the required memory into smaller blocks is
left a TODO for later (this can be useful for devices without IOMMU).

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[memoryplanner] Introducing memory planner
Parichay Kapoor [Tue, 10 Aug 2021 12:57:56 +0000 (21:57 +0900)]
[memoryplanner] Introducing memory planner

This patch introduces memory planner interface which will provide a
layout plan given the various memories required by the model.
Different memory allocations plans/strategies will be implemented by
extending this class with the given interface.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[manager] Weight/gradient allocation simplify
Parichay Kapoor [Tue, 10 Aug 2021 10:57:47 +0000 (19:57 +0900)]
[manager] Weight/gradient allocation simplify

Simplify weight and gradient allocation out to the most basic strategy
so that the memory optimization can be applied on it.
With this patch, the memory for gradient is not shared anymore till the
memory optimization is implemented.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[Dataset] Implement samplewise for file
Jihoon Lee [Tue, 10 Aug 2021 05:46:18 +0000 (14:46 +0900)]
[Dataset] Implement samplewise for file

**Changes proposed in this PR:**
- implements samplewise size(), finalize() for file
- add corresponding test

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI] Update description of ml_train_datagen_cb
Jihoon Lee [Thu, 19 Aug 2021 08:43:55 +0000 (17:43 +0900)]
[CAPI] Update description of ml_train_datagen_cb

This patch updates ml_train_datagen_cb with newly added api sets and
clarify the parameter @a last means.

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI] Reflect feature review
Jihoon Lee [Thu, 19 Aug 2021 02:28:44 +0000 (11:28 +0900)]
[CAPI] Reflect feature review

This patch reflects feature review.

**Changes Proposed**
- Adding Not Supported enum

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[Backbone/CAPI] Implement backbone layer creation
Jihoon Lee [Wed, 18 Aug 2021 07:32:26 +0000 (16:32 +0900)]
[Backbone/CAPI] Implement backbone layer creation

**Changes proposed in this PR:**
- Connect enum to nnstreamer backbone
- Change `model_file` property to `model_path` for generality
- Add handling for when model_file property is loaded from interpretation

Resolves #1072

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Cc: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[Dataset] Implement samplewise producer for random
Jihoon Lee [Mon, 9 Aug 2021 10:05:14 +0000 (19:05 +0900)]
[Dataset] Implement samplewise producer for random

This patch implements samplewise producer for random and its
corresponding test abstraction

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[Trivial] layer client test was not running
Jihoon Lee [Thu, 19 Aug 2021 02:24:11 +0000 (11:24 +0900)]
[Trivial] layer client test was not running

This patch changes test variable argument in order for layer client
tests to run

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[context/graph] Catch lifespan + 3-way execution order
Parichay Kapoor [Tue, 10 Aug 2021 10:21:27 +0000 (19:21 +0900)]
[context/graph] Catch lifespan + 3-way execution order

This patch adds the setup to catch lifespan set by the layer developer
when requesting for tensors.
Further, the execution order now consists of 3 values - fowarding,
calcGradient and calcDerivative. The memory optimization will now be
done using the 3 functions to be called in order.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[manager] Set the default usage timestamp for tensors
Parichay Kapoor [Wed, 4 Aug 2021 11:03:04 +0000 (20:03 +0900)]
[manager] Set the default usage timestamp for tensors

Set default usage time index for all the tensors based on when the layer
is to be run in the sorted list.
The usage time indexes are stored in each GraphNode by the GraphCore.
When creating the sorted list, each layer is updated to store its
order location/index when executed in this sequence for forwarding as
well as backwarding.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[layers] Make requested weight/tensor names unique
Parichay Kapoor [Wed, 4 Aug 2021 10:37:27 +0000 (19:37 +0900)]
[layers] Make requested weight/tensor names unique

Update the layers to make the requested weights and tensors names to be
unique. This done using the layer name with the postfix of the
weights/tensors name making the combination unqiue as layer name is
unique.

InitLayerContext is updated to provide access to the layer name while
finalizing the layer.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[graph/neuralnet] Move manager from neuralnet to netgraph
Parichay Kapoor [Wed, 4 Aug 2021 05:08:31 +0000 (14:08 +0900)]
[graph/neuralnet] Move manager from neuralnet to netgraph

Move manager from NeuralNetwork class to NetworkGraph class.
Some corresponding minor adjustments are also added.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[Dataset] Add interface for samplewise producer
Jihoon Lee [Mon, 9 Aug 2021 08:09:50 +0000 (17:09 +0900)]
[Dataset] Add interface for samplewise producer

This patch adds interface to migrate batchwise producer to samplewise
producer

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[Dataset] Add user_data set Property handler
Jihoon Lee [Thu, 15 Jul 2021 10:14:22 +0000 (19:14 +0900)]
[Dataset] Add user_data set Property handler

**Changes proposed in this PR:**
- PtrType properties
- Add user_data set property handler for backward competibility

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI] Reflect privilege reviews
Jihoon Lee [Fri, 6 Aug 2021 08:33:13 +0000 (17:33 +0900)]
[CAPI] Reflect privilege reviews

This patch update document to add privilege to `ml_train_model_save()`
and `ml_train_model_load()`

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[fix] ahub issues
hyeonseok lee [Tue, 17 Aug 2021 11:20:51 +0000 (20:20 +0900)]
[fix] ahub issues

 - Added try catch statement
 - Check value to avoid divide by zero
 - Added const keyword to match override function
 - Initialize member variable in constructor

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years ago[graph/neuralnet] Move manager from neuralnet to netgraph
Parichay Kapoor [Wed, 4 Aug 2021 05:08:31 +0000 (14:08 +0900)]
[graph/neuralnet] Move manager from neuralnet to netgraph

Move manager from NeuralNetwork class to NetworkGraph class.
Some corresponding minor adjustments are also added.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[graph] Update check for input node
Parichay Kapoor [Wed, 4 Aug 2021 03:54:13 +0000 (12:54 +0900)]
[graph] Update check for input node

Update the check for checking if a given node is input node.
The check uses in-degree for the node to be 0.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[graphnode] Remove index from graph node
Parichay Kapoor [Wed, 4 Aug 2021 03:46:41 +0000 (12:46 +0900)]
[graphnode] Remove index from graph node

Currently each graph node has two identifiers - name and idx - where
both were unique. This patch removes the idx as the indentifier.
The corresponding functions are also removed.
In order to support fast fetching of nodes, graphcore has a map nodes.

This change revealed a bug with the names in the code. As ini does not
allow case sensitive names, the names from the ini are changed to lower
case. However, input_layers property names were allowed to be
case-sensitive. Further names set by the API were also case sensitive.
This patch forces all such names to be case-insensitive to lowercase.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[graph] Remove layerv1 initialize code
Parichay Kapoor [Wed, 4 Aug 2021 02:07:31 +0000 (11:07 +0900)]
[graph] Remove layerv1 initialize code

Remove layerv2 initialize code for network graph.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[CAPI] Assign values to the enum
Jihoon Lee [Fri, 13 Aug 2021 02:37:58 +0000 (11:37 +0900)]
[CAPI] Assign values to the enum

This patch assigns values to enums in nntrainer-api-common.h

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI] Update regarding the review script
Jihoon Lee [Fri, 6 Aug 2021 08:33:13 +0000 (17:33 +0900)]
[CAPI] Update regarding the review script

This patch updates nntrainer.h header with regard to the review script

- @a -> @c
- merge conflict with usage / mode

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[Trivial] dataset usage enum rename
Jihoon Lee [Thu, 15 Jul 2021 07:50:58 +0000 (16:50 +0900)]
[Trivial] dataset usage enum rename

This patch renames enumeration as discussed ahead of time

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI] Add permission denied return value
Jihoon Lee [Fri, 6 Aug 2021 08:54:48 +0000 (17:54 +0900)]
[CAPI] Add permission denied return value

As privilege is sometimes needed by dataset, add permission denied for
`ml_train_dataset_add_file()` and `ml_train_model_construct()`

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI] Change privilege to remark
Jihoon Lee [Fri, 6 Aug 2021 07:43:49 +0000 (16:43 +0900)]
[CAPI] Change privilege to remark

This patch change privilege to remark as recommended

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI] Add privilege to documentation
Jihoon Lee [Thu, 5 Aug 2021 03:55:33 +0000 (12:55 +0900)]
[CAPI] Add privilege to documentation

This patch adds privilege to documentation as guided

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI/ACR] Fix typo and doxygen tag
Jihoon Lee [Tue, 3 Aug 2021 04:36:56 +0000 (13:36 +0900)]
[CAPI/ACR] Fix typo and doxygen tag

This patch reflects review from #1371, #1446

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[Trivial] Reflect acr review script
Jihoon Lee [Tue, 3 Aug 2021 04:14:01 +0000 (13:14 +0900)]
[Trivial] Reflect acr review script

This patch reflect acr review scripts, which contains some trivial fixes
like adding a period at the end.

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI/ACR] Add remark on tensor info getter
Jihoon Lee [Tue, 3 Aug 2021 03:33:51 +0000 (12:33 +0900)]
[CAPI/ACR] Add remark on tensor info getter

This patch adds remark on tensor info getter that info should not be
destroyed when fail to create

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[fixes] Add bug fixes from asan
Jihoon Lee [Thu, 12 Aug 2021 08:19:50 +0000 (17:19 +0900)]
[fixes] Add bug fixes from asan

**Changes proposed in this PR:**
- Fix issues from address sanitizer
- Enable deprecated-declaration warning but disable for some files

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[fixes] Add bug fixes found with address sanitizer
Parichay Kapoor [Thu, 12 Aug 2021 06:44:25 +0000 (15:44 +0900)]
[fixes] Add bug fixes found with address sanitizer

This patch adds bug fixes after adding fixes related to address
sanitzer:
- NNStreamerLayer wasnt freeing resources upon destruction
- accessing released model handle in capi unittest
- not releasing model handle before end of unittest

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[Dataset] Remove unused part from refactoring
Jihoon Lee [Wed, 14 Jul 2021 16:42:49 +0000 (01:42 +0900)]
[Dataset] Remove unused part from refactoring

This patch removes parts that are no longer used from the refactoring

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[Dataset] Adopt new datagetter to databuffer
Jihoon Lee [Wed, 14 Jul 2021 15:43:32 +0000 (00:43 +0900)]
[Dataset] Adopt new datagetter to databuffer

**Changes Propsed in this PR**
1. Adopt new data getter to neuralnet / factory
2. Fix: bugs found in raw_file_data_producer
3. Fix: databuffer not feeding last iteration when error

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[tensor/test] Unittests for tensor allocate and initialize
Parichay Kapoor [Tue, 10 Aug 2021 01:29:31 +0000 (10:29 +0900)]
[tensor/test] Unittests for tensor allocate and initialize

Add unittests for tensor allocation and initialize the tensor memory in
various scenarios.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[layer] Add support for initializer in requestTensor
Parichay Kapoor [Tue, 27 Jul 2021 08:08:46 +0000 (17:08 +0900)]
[layer] Add support for initializer in requestTensor

This patch adds support for initializer in the requestTensor interface
of the layer context.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[weight] Weight cleanup related to initializer
Parichay Kapoor [Tue, 27 Jul 2021 06:47:48 +0000 (15:47 +0900)]
[weight] Weight cleanup related to initializer

This patch provides weight cleanup related to the initializer.
Weights dont take care of the initializer but rather let var_grad handle
it which in turn let tensor handle it.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[var_grad] Support initializer for var_grad
Parichay Kapoor [Tue, 27 Jul 2021 06:36:23 +0000 (15:36 +0900)]
[var_grad] Support initializer for var_grad

Support initializer for variables and gradient in var_grad with the
initializer of the tensor. This allows initialization to be done for
requested tensors as well, which will be updated in a later commit.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[Tensor] Support initializer with tensor
Parichay Kapoor [Tue, 27 Jul 2021 06:09:32 +0000 (15:09 +0900)]
[Tensor] Support initializer with tensor

Support initializer with tensor object natively.
Initialization will be done when the memory is allocated or can be
invoked manually as well.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[tensor] WeightInitializer refactor to Tensor::Initializer
Parichay Kapoor [Tue, 27 Jul 2021 05:38:10 +0000 (14:38 +0900)]
[tensor] WeightInitializer refactor to Tensor::Initializer

Rename/cleanup tensor functions to adhere with STL like function names.
Further, weight initializer has officially become tensor initializer.
The usage will be updated soon.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[tensor/weight] Update WeightInitializer to TensorInitializer
Parichay Kapoor [Mon, 26 Jul 2021 12:16:07 +0000 (21:16 +0900)]
[tensor/weight] Update WeightInitializer to TensorInitializer

Update weightInitializer to TensorInitializer.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[Dataset] Change factory signature
Jihoon Lee [Wed, 14 Jul 2021 09:58:22 +0000 (18:58 +0900)]
[Dataset] Change factory signature

Add properties to factory, given contructor argument, it is delegated to
the constructor

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Cc: Parichay Kapoor<pk.kapoor@samsung.com>
Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[devel/pkg] Update the devel packaging
Parichay Kapoor [Mon, 26 Jul 2021 08:47:14 +0000 (17:47 +0900)]
[devel/pkg] Update the devel packaging

Update the devel packaging to remove unnecessary headers.
- weight/var_grad headers are removed
- app_context and other graph, network related headers are removed. They
can be added once network/neural network headers are refactored for
custom models.
- manager and dynamic_training_optimization header is removed
- databuffer headers are commented. They will be updated once dataset
refactoring is done.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[unittest] enable commented unittest
hyeonseok lee [Fri, 6 Aug 2021 03:51:40 +0000 (12:51 +0900)]
[unittest] enable commented unittest

 - Using ValuesIn function instead of Values cause the arguments are more than 50
 - Handle unused variables

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years ago[tensordim] move implementation to source file
hyeonseok lee [Fri, 30 Jul 2021 12:31:10 +0000 (21:31 +0900)]
[tensordim] move implementation to source file

 - header file remains without implementaion

Build test: [X]Passed [ ]Failed [ ]Skipped
Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years agoDebian: you don't have debian-build meson option.
MyungJoo Ham [Thu, 5 Aug 2021 06:58:31 +0000 (15:58 +0900)]
Debian: you don't have debian-build meson option.

Remove unsupported meson build option from debian script.

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
3 years ago[mesonbuild] update meson build
hyeonseok lee [Wed, 4 Aug 2021 04:07:59 +0000 (13:07 +0900)]
[mesonbuild] update meson build

 - Make platform meson option
 - Check sysroot

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years ago[optimizer] Rename initialize to finalize
Parichay Kapoor [Wed, 4 Aug 2021 04:04:02 +0000 (13:04 +0900)]
[optimizer] Rename initialize to finalize

Rename initialize to finalize.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[layer] Update layer.h setProperty interface
Parichay Kapoor [Fri, 30 Jul 2021 05:37:45 +0000 (14:37 +0900)]
[layer] Update layer.h setProperty interface

Update the interface for the layer.h setProperty.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[optimizer] Cleanup optimizer and remove dependency on weight
Parichay Kapoor [Fri, 30 Jul 2021 05:26:25 +0000 (14:26 +0900)]
[optimizer] Cleanup optimizer and remove dependency on weight

Cleanup optimizer interface where some extra interfaces have been
removed, and arguments interface has been updated.
Further, the dependency on weight header has been removed, and added an
optimizer context interface which provides an interface to weight
related tensors required by the optimizer.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[Dataset] Implement data getter
Jihoon Lee [Tue, 13 Jul 2021 16:14:28 +0000 (01:14 +0900)]
[Dataset] Implement data getter

This patch implements data getter.
There are two types of data getter.

1. queue multiple iterations and get one by one
(`DataBuffer::startFetchWorker()`, `DataBuffer::fetch()`)

2. generate one iteration on the fly (`DataBuffer::batcher()`)

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[tensorfilter] Update tensor filter to use ccapi
Parichay Kapoor [Tue, 3 Aug 2021 03:33:49 +0000 (12:33 +0900)]
[tensorfilter] Update tensor filter to use ccapi

Update nntrainer's tensor filter for nnstreamer to depend only on ccapi
and nntrainer directly and internal headers.

See Also #986

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[DataSet] Refactor to use properties
Jihoon Lee [Tue, 13 Jul 2021 13:29:51 +0000 (22:29 +0900)]
[DataSet] Refactor to use properties

This patch enables properties in dataset while changing signature of
dataset

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[Dataset] Add thread safe queue for the buffer
Jihoon Lee [Tue, 13 Jul 2021 11:18:06 +0000 (20:18 +0900)]
[Dataset] Add thread safe queue for the buffer

Currently, databuffer is not queuing any iteration but requesting buffer
on-demand basis.
This patch adds thread safe queue for the buffer to queue multiple
buffers

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[weight/var_grad] Remove exposure of weight/var_grad
Parichay Kapoor [Mon, 26 Jul 2021 08:27:06 +0000 (17:27 +0900)]
[weight/var_grad] Remove exposure of weight/var_grad

This patch updates the usage of weight and var_grad headers in order to
hide them internally and not expose them in the devel headers.
The below changes are made to support this:
- weight and var_grad specs are declared separately in tensor wrapper
specs header.
- layer_context is made indepedent of the weight and var_grad
definition. The usages and implementation are moved to layer_context
souce file.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[utils] Split parse_utils into two for cleaner headers
Parichay Kapoor [Mon, 26 Jul 2021 06:58:47 +0000 (15:58 +0900)]
[utils] Split parse_utils into two for cleaner headers

Carve props_utils out of parse_utils to get cleaner headers
being shipped out with devel package.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[optimizer] Deprecate optimizer factory
Parichay Kapoor [Mon, 26 Jul 2021 05:49:47 +0000 (14:49 +0900)]
[optimizer] Deprecate optimizer factory

Deprecate usage of optimizer factory and replace it with app_context.
Further, app_context throw code update from runtime_error to
invalid_argument when the requested type of object is not found.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[test] capi unittest fix
Parichay Kapoor [Tue, 3 Aug 2021 02:28:08 +0000 (11:28 +0900)]
[test] capi unittest fix

C-API unittest was checking for throw but instead should be checking for
return value. This patch fixes the test.
Further, the unittests tries to free unallocated tensordim in error
cases. This patch resolves it too.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[CAPI] Move save/load api to public
Jihoon Lee [Mon, 26 Jul 2021 09:14:40 +0000 (18:14 +0900)]
[CAPI] Move save/load api to public

This patch moves save/load api to public

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[tensor_filter] Remove dependency on nntrainer::Tensor
Parichay Kapoor [Mon, 26 Jul 2021 11:40:37 +0000 (20:40 +0900)]
[tensor_filter] Remove dependency on nntrainer::Tensor

Remove tensor filter implementations dependency on nntrianer::Tensor.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[tensor filter] Bugfix for nnstreamer's tensor filter
Parichay Kapoor [Mon, 26 Jul 2021 10:44:50 +0000 (19:44 +0900)]
[tensor filter] Bugfix for nnstreamer's tensor filter

This patch adds bugfix for the nnstreamer's tensorfilter of nntrainer.
- include proper headers required by the tensor filter header
- use the correct offset for the multi-input scenario

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[c++/model] Add model inference to c++ API
Parichay Kapoor [Mon, 26 Jul 2021 10:34:02 +0000 (19:34 +0900)]
[c++/model] Add model inference to c++ API

Add model inference to the c++ API using the vector of float ptrs as
inputs and outputs.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[loss] BugFix for regularization loss
Parichay Kapoor [Mon, 2 Aug 2021 08:55:05 +0000 (17:55 +0900)]
[loss] BugFix for regularization loss

This patch adds bugfix for regularization loss calculation.
Regularization loss must be calculated before the weight is updated in
backwarding, but this was not being done.
This patch provides the corresponding fix.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[test] Enable c/cc diabled tests
Parichay Kapoor [Fri, 30 Jul 2021 07:11:08 +0000 (16:11 +0900)]
[test] Enable c/cc diabled tests

This patch enables c/cc disabled tests.
Corresponding required updates in the tests are also added.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[dataset] Add file producer
Jihoon Lee [Mon, 12 Jul 2021 12:28:12 +0000 (21:28 +0900)]
[dataset] Add file producer

This patch adds file producer which abstracts reading a raw file. Also,
this component makes sure that the file accepts various kind of input
shapes.

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI] Implement ml_train_model_get_input|output_dims
hyeonseok lee [Wed, 21 Jul 2021 11:22:30 +0000 (20:22 +0900)]
[CAPI] Implement ml_train_model_get_input|output_dims

 - Implement ml_train_model_get_input_dimensions
 - Implement ml_train_model_get_output_dimensions
 - Make a unittest for implemented apis

Self evaluation:

Build test: [X]Passed [ ]Failed [ ]Skipped
Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years ago[tensor_dim] package tensor_dim.h with ccapi
hyeonseok lee [Thu, 29 Jul 2021 10:47:44 +0000 (19:47 +0900)]
[tensor_dim] package tensor_dim.h with ccapi

 - Package tensor_dim with ccapi
 - MAXDIM belongs to tensorDim class

Self evaluation:

Build test: [X]Passed [ ]Failed [ ]Skipped
Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years ago[CAPI] Expose layer enums
Jihoon Lee [Tue, 27 Jul 2021 02:41:07 +0000 (11:41 +0900)]
[CAPI] Expose layer enums

This patch contains layer enums to be exposed

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI] Change tensor dim -> information
Jihoon Lee [Wed, 28 Jul 2021 03:11:26 +0000 (12:11 +0900)]
[CAPI] Change tensor dim -> information

This patch updates description of the tensors_info, while explicitly
stating that the information object is newly created

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Cc: Hyeonseok Lee<hs89.lee@samsung.com>
Cc: Parichay Kapoor<pk.kapoor@samsung.com>
Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[ Layer ] Change OutputLayer to MultiOutLayer
jijoong.moon [Wed, 28 Jul 2021 02:41:10 +0000 (11:41 +0900)]
[ Layer ] Change OutputLayer to MultiOutLayer

Currently the feature of OutputLayer is not match. It is more like
Multiout layer. The type name and others are already multiout.

This commit includes,
  . Class name changed : OutputLayer --> MultiOutLayer
  . File name changed : output_layer.x --> multiout_layer.x

**Self evaluation:**
1. Build test:  [X]Passed [ ]Failed [ ]Skipped
2. Run test:  [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: jijoong.moon <jijoong.moon@samsung.com>
3 years ago[CAPI] Add ml_train_model_get_layer to internal
Jihoon Lee [Mon, 26 Jul 2021 09:08:08 +0000 (18:08 +0900)]
[CAPI] Add ml_train_model_get_layer to internal

This patch adds ml_train_model_get_layer to internal api

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[CAPI] Deprecate old dataset apis
Jihoon Lee [Mon, 26 Jul 2021 08:35:32 +0000 (17:35 +0900)]
[CAPI] Deprecate old dataset apis

This patch updates to deprecate old dataset apis

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[Dataset] Add callback generator
Jihoon Lee [Mon, 12 Jul 2021 09:36:49 +0000 (18:36 +0900)]
[Dataset] Add callback generator

This patch add callback producer to abstract generator

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[ GRU ] Add dropout mask
jijoong.moon [Fri, 23 Jul 2021 00:02:14 +0000 (09:02 +0900)]
[ GRU ] Add dropout mask

This commit includes:
. Implementation of drop out mask

**Self evaluation:**
1. Build test:  [X]Passed [ ]Failed [ ]Skipped
2. Run test:  [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: jijoong.moon <jijoong.moon@samsung.com>
3 years ago[Dataset] Add test for the random dataproducer
Jihoon Lee [Mon, 12 Jul 2021 06:47:44 +0000 (15:47 +0900)]
[Dataset] Add test for the random dataproducer

**Changes proposed in this PR:**
- Add random data producer tests
- Add abstract test for more data producers

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[ RNNT LOSS ] Add skeleton code for rnnt loss
jijoong.moon [Thu, 22 Jul 2021 06:35:30 +0000 (15:35 +0900)]
[ RNNT LOSS ] Add skeleton code for rnnt loss

This commit includes:
  . Sekeleton Code for RNNT Loss as an Custom Loss Layer
  . Sementic Unit Test

**Self evaluation:**
1. Build test:  [X]Passed [ ]Failed [ ]Skipped
2. Run test:  [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: jijoong.moon <jijoong.moon@samsung.com>
3 years ago[Android] Fix undefined definition
Jihoon Lee [Mon, 26 Jul 2021 10:47:28 +0000 (19:47 +0900)]
[Android] Fix undefined definition

This patch defines ML_API_COMMON_ROOT previously undefined

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[ LSTM ] Add dropout mask
jijoong.moon [Thu, 22 Jul 2021 23:55:44 +0000 (08:55 +0900)]
[ LSTM ] Add dropout mask

In this commit,
. drop out mask is added.
. calculation of drop out is implemented.

**Self evaluation:**
1. Build test:  [X]Passed [ ]Failed [ ]Skipped
2. Run test:  [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: jijoong.moon <jijoong.moon@samsung.com>
3 years ago[ RNN ] Add dropout mask
jijoong.moon [Thu, 22 Jul 2021 23:39:49 +0000 (08:39 +0900)]
[ RNN ] Add dropout mask

This commit includes :
  . implementation of mask for dropout

**Self evaluation:**
1. Build test:  [X]Passed [ ]Failed [ ]Skipped
2. Run test:  [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: jijoong.moon <jijoong.moon@samsung.com>
3 years ago[Android] Add missing dependency
Jihoon Lee [Mon, 26 Jul 2021 05:45:57 +0000 (14:45 +0900)]
[Android] Add missing dependency

This patch adds mssing dependency to the android application build

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Jihoon Lee <jhoon.it.lee@samsung.com>
3 years ago[ Fix ] Fix android build error with meson
jijoong.moon [Mon, 26 Jul 2021 01:27:56 +0000 (10:27 +0900)]
[ Fix ] Fix android build error with meson

This commit includes,
  . Fix android build issues
    : openblas link and download

**Self evaluation:**
1. Build test:  [X]Passed [ ]Failed [ ]Skipped
2. Run test:  [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: jijoong.moon <jijoong.moon@samsung.com>
3 years ago[Fix] Initialize member variable
hyeonseok lee [Thu, 22 Jul 2021 04:33:17 +0000 (13:33 +0900)]
[Fix] Initialize member variable

 - Initialize std::function variable in constructor

Self evaluation:

Build test: [X]Passed [ ]Failed [ ]Skipped
Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years ago[ TEST ] fix compilation error
jijoong.moon [Thu, 22 Jul 2021 13:03:51 +0000 (22:03 +0900)]
[ TEST ] fix compilation error

This commit includes,
  . compilation error fix : unused + wrong type comparison

**Self evaluation:**
1. Build test:  [X]Passed [ ]Failed [ ]Skipped
2. Run test:  [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: jijoong.moon <jijoong.moon@samsung.com>
3 years ago[unittest] multi output unittest
hyeonseok lee [Thu, 8 Jul 2021 10:32:09 +0000 (19:32 +0900)]
[unittest] multi output unittest

 - Modify unittest to test multi output case
 - Make setLabel function in neuralnet
   to set labels at all loss layers

Resolve #1367

**Self evaluation:**
1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years ago[network_graph] revise addlosslayer for multi output
hyeonseok lee [Fri, 25 Jun 2021 12:32:48 +0000 (21:32 +0900)]
[network_graph] revise addlosslayer for multi output

 - reorder addlosslayer to run before topologicalsort
 - Replace node_list rather than delete when the loss is entropy

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years ago[graph_core] Implement input_list, output_list for multi input, output
hyeonseok lee [Fri, 25 Jun 2021 06:39:53 +0000 (15:39 +0900)]
[graph_core] Implement input_list, output_list for multi input, output

- Make input_list, output_list and its getter to support multi input, output

Signed-off-by: hyeonseok lee <hs89.lee@samsung.com>
3 years ago[layer] Update dropout layer to V2
Parichay Kapoor [Wed, 21 Jul 2021 07:47:32 +0000 (16:47 +0900)]
[layer] Update dropout layer to V2

Update dropout layer to V2 design.

Signed-off-by: Parichay Kapoor <kparichay@gmail.com>
3 years ago[coverity] Fixes realtd to coverity
Parichay Kapoor [Wed, 21 Jul 2021 03:18:46 +0000 (12:18 +0900)]
[coverity] Fixes realtd to coverity

Add fixes related to coverity.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[svace] Fix svace issues
Parichay Kapoor [Tue, 20 Jul 2021 11:15:50 +0000 (20:15 +0900)]
[svace] Fix svace issues

Fix svace issues for layer_v2 branch

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[layer_v2] Merge commit for branch layer_v2
Parichay Kapoor [Tue, 20 Jul 2021 09:51:19 +0000 (18:51 +0900)]
[layer_v2] Merge commit for branch layer_v2

This commit forms the merge commit including minor updates
while rebase layer_v2 with main branch so as to apply it.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[LayerV1] Delete for LayerV1
Parichay Kapoor [Thu, 15 Jul 2021 02:05:59 +0000 (11:05 +0900)]
[LayerV1] Delete for LayerV1

This patch deletes LayerV1 headers and its implementations.
Some of the relevant codes from LayerV1 are moved to either LayerNode or
to LayerDevel.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[layernode] Add print functionality to LayerNode
Parichay Kapoor [Wed, 14 Jul 2021 12:48:44 +0000 (21:48 +0900)]
[layernode] Add print functionality to LayerNode

Add print functionality from LayerV1 to LayerNode.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[layer] Update NNStreamer layer to V2
Parichay Kapoor [Wed, 14 Jul 2021 12:10:50 +0000 (21:10 +0900)]
[layer] Update NNStreamer layer to V2

Update NNStreamer layer to V2 design.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>
3 years ago[layer] Enable permute layer for V2
Parichay Kapoor [Wed, 14 Jul 2021 12:03:22 +0000 (21:03 +0900)]
[layer] Enable permute layer for V2

Enable permute layer for V2 design.

Signed-off-by: Parichay Kapoor <pk.kapoor@samsung.com>