Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[onert] Add unit tests for training Softmax op #13197

Merged
merged 2 commits into from
Jun 19, 2024
Merged

[onert] Add unit tests for training Softmax op #13197

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d2b7918
[onert] Add unit tests for training Softmax op
ragmani Jun 14, 2024
5dc9e35
Add calling `cgen.markAllOpsAsTrainable()`
ragmani Jun 17, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
102 changes: 102 additions & 0 deletions tests/nnfw_api/src/GenModelTests/nontrainable_op_trains/Softmax.test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,102 @@
/*
* Copyright (c) 2024 Samsung Electronics Co., Ltd. All Rights Reserved
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/

#include "GenModelTrain.h"

TEST_F(GenModelTrain, NonTrainableOps_FC_Softmax)
{
// (( Input 0 )) -> [ FC ] -> [ Softmax ] -> (( Output 0 ))
{
CirclePlusGen cgen;

uint32_t weight_buf = cgen.addBuffer(std::vector<float>(8 * 2, 0.f));
uint32_t bias_buf = cgen.addBuffer(std::vector<float>(8, 0.f));
int input = cgen.addTensor({{1, 2}, circle::TensorType::TensorType_FLOAT32});
int weight = cgen.addTensor({{8, 2}, circle::TensorType::TensorType_FLOAT32, weight_buf});
int bias = cgen.addTensor({{8}, circle::TensorType::TensorType_FLOAT32, bias_buf});
int fc_output = cgen.addTensor({{1, 8}, circle::TensorType::TensorType_FLOAT32});
int output = cgen.addTensor({{1, 8}, circle::TensorType::TensorType_FLOAT32});
const float beta = 1.0f;
cgen.addOperatorFullyConnected({{input, weight, bias}, {fc_output}});
cgen.addOperatorSoftmax({{fc_output}, {output}}, beta);
cgen.setInputsAndOutputs({input}, {output});

float learning_rate = 0.01f;
int32_t batch_size = 1;
cgen.addTrainInfo({circle::Optimizer::Optimizer_SGD, learning_rate,
circle::LossFn::LossFn_MEAN_SQUARED_ERROR,
circle::LossReductionType::LossReductionType_SumOverBatchSize, batch_size});
ragmani marked this conversation as resolved.
Show resolved Hide resolved
cgen.markAllOpsAsTrainable();

_context = std::make_unique<GenModelTrainContext>(cgen.finish());
_context->addTrainCase(
uniformTCD<float>({{{1, 3}}, {{2, 1}}}, // inputs
{{{0, 1, 0, 0, 0, 0, 0, 0}}, {{0, 0, 0, 0, 0, 1, 0, 0}}}, // expected
{0.1092f} // loss
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The result of executing the same model with the same data on tensorflow:

Epoch 1/5
2/2 [==============================] - 0s 3ms/step - loss: 0.1094 - categorical_accuracy: 0.0000e+00
Epoch 2/5
2/2 [==============================] - 0s 1ms/step - loss: 0.1093 - categorical_accuracy: 0.5000
Epoch 3/5
2/2 [==============================] - 0s 1ms/step - loss: 0.1092 - categorical_accuracy: 0.5000
Epoch 4/5
2/2 [==============================] - 0s 1ms/step - loss: 0.1092 - categorical_accuracy: 0.5000
Epoch 5/5
2/2 [==============================] - 0s 1ms/step - loss: 0.1091 - categorical_accuracy: 0.5000

));

_context->setBackends({"train"});
_context->setEpoch(4);

SUCCEED();
}
}

TEST_F(GenModelTrain, neg_NonTrainableOps_Softmax_InvalidShape)
{
CirclePlusGen cgen;

int input = cgen.addTensor({{2, 1}, circle::TensorType::TensorType_FLOAT32});
// Invalid shape: output shape should be equal to input shape
int output = cgen.addTensor({{2, 2}, circle::TensorType::TensorType_FLOAT32});
const float beta = 1.0f;
cgen.addOperatorSoftmax({{input}, {output}}, beta);
cgen.setInputsAndOutputs({input}, {output});

float learning_rate = 0.01f;
int32_t batch_size = 1;
cgen.addTrainInfo({circle::Optimizer::Optimizer_SGD, learning_rate,
circle::LossFn::LossFn_MEAN_SQUARED_ERROR,
circle::LossReductionType::LossReductionType_SumOverBatchSize, batch_size});
ragmani marked this conversation as resolved.
Show resolved Hide resolved
cgen.markAllOpsAsTrainable();

_context = std::make_unique<GenModelTrainContext>(cgen.finish());
_context->setBackends({"train"});
_context->expectFailCompile();
}

TEST_F(GenModelTrain, neg_NonTrainableOps_Softmax_InvalidType)
{
CirclePlusGen cgen;

// Invalid type: input tensor type should be FLOAT32
int input = cgen.addTensor({{2, 2}, circle::TensorType::TensorType_INT32});
int output = cgen.addTensor({{2, 2}, circle::TensorType::TensorType_FLOAT32});
const float beta = 1.0f;
cgen.addOperatorSoftmax({{input}, {output}}, beta);
cgen.setInputsAndOutputs({input}, {output});

float learning_rate = 0.01f;
int32_t batch_size = 1;
cgen.addTrainInfo({circle::Optimizer::Optimizer_SGD, learning_rate,
circle::LossFn::LossFn_MEAN_SQUARED_ERROR,
circle::LossReductionType::LossReductionType_SumOverBatchSize, batch_size});
ragmani marked this conversation as resolved.
Show resolved Hide resolved
cgen.markAllOpsAsTrainable();

_context = std::make_unique<GenModelTrainContext>(cgen.finish());
_context->setBackends({"train"});
_context->expectFailModelLoad();
}
Loading