Moved examples to examples/torch dir and tests to tests/torch dir (#725)

* Moved examples to examples/torch dir * Moved tests to tests/torch dir * Update paths * Fixed tests * Added init
openvinotoolkit · Jun 3, 2021 · aafe4bf · aafe4bf
1 parent 7e77f7d
commit aafe4bf
Show file tree

Hide file tree

Showing 771 changed files with 513 additions and 482 deletions.
diff --git a/.gitignore b/.gitignore
@@ -108,4 +108,4 @@ ENV/
 *.tar
 
 # object detection eval results
-examples/object_detection/eval/
+examples/torch/object_detection/eval/
diff --git a/README.md b/README.md
@@ -65,9 +65,9 @@ For more details about the framework architecture, refer to the [NNCFArchitectur
 For a quicker start with NNCF-powered compression, you can also try the sample scripts, each of which provides a basic training pipeline for classification, semantic segmentation and object detection neural network training correspondingly.
 
 To run the samples please refer to the corresponding tutorials:
-- [Image Classification sample](examples/classification/README.md)
-- [Object Detection sample](examples/object_detection/README.md)
-- [Semantic Segmentation sample](examples/semantic_segmentation/README.md)
+- [Image Classification sample](examples/torch/classification/README.md)
+- [Object Detection sample](examples/torch/object_detection/README.md)
+- [Semantic Segmentation sample](examples/torch/semantic_segmentation/README.md)
 
 ## Third-party repository integration
 NNCF may be straightforwardly integrated into training/evaluation pipelines of third-party repositories.
@@ -116,7 +116,7 @@ Use one of the Dockerfiles in the [docker](./docker) directory to build an image
 
 **NOTE**: If you want to use sample training scripts provided in the NNCF repository under `examples`, you should install the corresponding Python package dependencies:
 ```
-pip install -r examples/requirements.txt
+pip install -r examples/torch/requirements.txt
 ```
 
 ## Contributing

diff --git a/docs/compression_algorithms/Quantization.md b/docs/compression_algorithms/Quantization.md
@@ -223,7 +223,7 @@ For automatic mixed-precision selection it's recommended to use the following te
 
 Note, optimizer parameters are model specific, this template contains optimal ones for ResNet-like models.
 
-Here's an [example](../../examples/classification/configs/mixed_precision/squeezenet1_1_imagenet_mixed_int_hawq.json) of 
+Here's an [example](../../examples/torch/classification/configs/mixed_precision/squeezenet1_1_imagenet_mixed_int_hawq.json) of
 using the template in the full configuration file.
 
 This template uses `plateau` scheduler. Though it usually leads to a lot of epochs of tuning for achieving a good 

diff --git a/examples/common/models/__init__.py b/examples/common/models/__init__.py
diff --git a/examples/__init__.py → examples/torch/__init__.py b/examples/__init__.py → examples/torch/__init__.py
diff --git a/examples/classification/README.md → examples/torch/classification/README.md b/examples/classification/README.md → examples/torch/classification/README.md
@@ -16,7 +16,7 @@ This sample demonstrates a DL model compression in case of an image-classificati
 To work with the sample you should install the corresponding Python package dependencies
 
 ```
-pip install -r examples/requirements.txt
+pip install -r examples/torch/requirements.txt
 ```
 
 ## Quantize FP32 Pretrained Model
@@ -30,7 +30,7 @@ To prepare the ImageNet dataset, refer to the following [tutorial](https://githu
 #### Run Classification Sample
 
 - If you did not install the package, add the repository root folder to the `PYTHONPATH` environment variable.
-- Go to the `examples/classification` folder.
+- Go to the `examples/torch/classification` folder.
 - Run the following command to start compression with fine-tuning on GPUs:
     ```
     python main.py -m train --config configs/quantization/mobilenet_v2_imagenet_int8.json --data /data/imagenet/ --log-dir=../../results/quantization/mobilenet_v2_int8/
@@ -88,7 +88,7 @@ To export a model to the OpenVINO IR and run it using the Intel® Deep Learning
 
 #### Binarization
 
-As an example of NNCF convolution binarization capabilities, you may use the configs in `examples/classification/configs/binarization` to binarize ResNet18. Use the same steps/command line parameters as for quantization (for best results, specify `--pretrained`), except for the actual binarization config path.
+As an example of NNCF convolution binarization capabilities, you may use the configs in `examples/torch/classification/configs/binarization` to binarize ResNet18. Use the same steps/command line parameters as for quantization (for best results, specify `--pretrained`), except for the actual binarization config path.
 
 ### Results for binarization
 |Model|Compression algorithm|Dataset|PyTorch compressed accuracy|NNCF config file|PyTorch Checkpoint|

diff --git a/examples/classification/__init__.py → examples/torch/classification/__init__.py b/examples/classification/__init__.py → examples/torch/classification/__init__.py
diff --git a/examples/classification/common.py → examples/torch/classification/common.py b/examples/classification/common.py → examples/torch/classification/common.py
@@ -1,6 +1,6 @@
 import torch.backends.cudnn as cudnn
 
-from examples.common.model_loader import load_resuming_model_state_dict_and_checkpoint_from_path
+from examples.torch.common.model_loader import load_resuming_model_state_dict_and_checkpoint_from_path
 from nncf.torch.utils import manual_seed
 
 

diff --git a/...nfigs/binarization/resnet18_imagenet.json → ...nfigs/binarization/resnet18_imagenet.json b/...nfigs/binarization/resnet18_imagenet.json → ...nfigs/binarization/resnet18_imagenet.json
diff --git a/...esnet18_imagenet_binarization_dorefa.json → ...esnet18_imagenet_binarization_dorefa.json b/...esnet18_imagenet_binarization_dorefa.json → ...esnet18_imagenet_binarization_dorefa.json
diff --git a/.../resnet18_imagenet_binarization_xnor.json → .../resnet18_imagenet_binarization_xnor.json b/.../resnet18_imagenet_binarization_xnor.json → .../resnet18_imagenet_binarization_xnor.json
diff --git a/...t_v2_imagenet_mixed_int_autoq_staged.json → ...t_v2_imagenet_mixed_int_autoq_staged.json b/...t_v2_imagenet_mixed_int_autoq_staged.json → ...t_v2_imagenet_mixed_int_autoq_staged.json
diff --git a/...mobilenet_v2_imagenet_mixed_int_hawq.json → ...mobilenet_v2_imagenet_mixed_int_hawq.json b/...mobilenet_v2_imagenet_mixed_int_hawq.json → ...mobilenet_v2_imagenet_mixed_int_hawq.json
diff --git a/...bilenet_v2_imagenet_mixed_int_manual.json → ...bilenet_v2_imagenet_mixed_int_manual.json b/...bilenet_v2_imagenet_mixed_int_manual.json → ...bilenet_v2_imagenet_mixed_int_manual.json
diff --git a/..._v2_imagenet_mixed_int_manual_staged.json → ..._v2_imagenet_mixed_int_manual_staged.json b/..._v2_imagenet_mixed_int_manual_staged.json → ..._v2_imagenet_mixed_int_manual_staged.json
diff --git a/...et50_imagenet_mixed_int_autoq_staged.json → ...et50_imagenet_mixed_int_autoq_staged.json b/...et50_imagenet_mixed_int_autoq_staged.json → ...et50_imagenet_mixed_int_autoq_staged.json
diff --git a/...ion/resnet50_imagenet_mixed_int_hawq.json → ...ion/resnet50_imagenet_mixed_int_hawq.json b/...ion/resnet50_imagenet_mixed_int_hawq.json → ...ion/resnet50_imagenet_mixed_int_hawq.json
diff --git a/...n/resnet50_imagenet_mixed_int_manual.json → ...n/resnet50_imagenet_mixed_int_manual.json b/...n/resnet50_imagenet_mixed_int_manual.json → ...n/resnet50_imagenet_mixed_int_manual.json
diff --git a/...t50_imagenet_mixed_int_manual_staged.json → ...t50_imagenet_mixed_int_manual_staged.json b/...t50_imagenet_mixed_int_manual_staged.json → ...t50_imagenet_mixed_int_manual_staged.json
diff --git a/...queezenet1_1_imagenet_mixed_int_hawq.json → ...queezenet1_1_imagenet_mixed_int_hawq.json b/...queezenet1_1_imagenet_mixed_int_hawq.json → ...queezenet1_1_imagenet_mixed_int_hawq.json
diff --git a/...eezenet1_1_imagenet_mixed_int_manual.json → ...eezenet1_1_imagenet_mixed_int_manual.json b/...eezenet1_1_imagenet_mixed_int_manual.json → ...eezenet1_1_imagenet_mixed_int_manual.json
diff --git a/...1_1_imagenet_mixed_int_manual_staged.json → ...1_1_imagenet_mixed_int_manual_staged.json b/...1_1_imagenet_mixed_int_manual_staged.json → ...1_1_imagenet_mixed_int_manual_staged.json
diff --git a/...n/configs/pruning/googlenet_imagenet.json → ...n/configs/pruning/googlenet_imagenet.json b/...n/configs/pruning/googlenet_imagenet.json → ...n/configs/pruning/googlenet_imagenet.json
diff --git a/...g/googlenet_pruning_geometric_median.json → ...g/googlenet_pruning_geometric_median.json b/...g/googlenet_pruning_geometric_median.json → ...g/googlenet_pruning_geometric_median.json
diff --git a/...ng/resnet18_pruning_geometric_median.json → ...ng/resnet18_pruning_geometric_median.json b/...ng/resnet18_pruning_geometric_median.json → ...ng/resnet18_pruning_geometric_median.json
diff --git a/...s/pruning/resnet18_pruning_magnitude.json → ...s/pruning/resnet18_pruning_magnitude.json b/...s/pruning/resnet18_pruning_magnitude.json → ...s/pruning/resnet18_pruning_magnitude.json
diff --git a/...on/configs/pruning/resnet34_imagenet.json → ...on/configs/pruning/resnet34_imagenet.json b/...on/configs/pruning/resnet34_imagenet.json → ...on/configs/pruning/resnet34_imagenet.json
diff --git a/...ng/resnet34_pruning_geometric_median.json → ...ng/resnet34_pruning_geometric_median.json b/...ng/resnet34_pruning_geometric_median.json → ...ng/resnet34_pruning_geometric_median.json
diff --git a/...ng/resnet50_pruning_geometric_median.json → ...ng/resnet50_pruning_geometric_median.json b/...ng/resnet50_pruning_geometric_median.json → ...ng/resnet50_pruning_geometric_median.json
diff --git a/...net161_imagenet_custom_quant_pattern.json → ...net161_imagenet_custom_quant_pattern.json b/...net161_imagenet_custom_quant_pattern.json → ...net161_imagenet_custom_quant_pattern.json
diff --git a/...s/quantization/inception_v3_imagenet.json → ...s/quantization/inception_v3_imagenet.json b/...s/quantization/inception_v3_imagenet.json → ...s/quantization/inception_v3_imagenet.json
diff --git a/...ntization/inception_v3_imagenet_int8.json → ...ntization/inception_v3_imagenet_int8.json b/...ntization/inception_v3_imagenet_int8.json → ...ntization/inception_v3_imagenet_int8.json
diff --git a/...s/quantization/mobilenet_v2_cifar100.json → ...s/quantization/mobilenet_v2_cifar100.json b/...s/quantization/mobilenet_v2_cifar100.json → ...s/quantization/mobilenet_v2_cifar100.json
diff --git a/...tion/mobilenet_v2_cifar100_asym_int8.json → ...tion/mobilenet_v2_cifar100_asym_int8.json b/...tion/mobilenet_v2_cifar100_asym_int8.json → ...tion/mobilenet_v2_cifar100_asym_int8.json
diff --git a/...s/quantization/mobilenet_v2_imagenet.json → ...s/quantization/mobilenet_v2_imagenet.json b/...s/quantization/mobilenet_v2_imagenet.json → ...s/quantization/mobilenet_v2_imagenet.json
diff --git a/...ntization/mobilenet_v2_imagenet_int8.json → ...ntization/mobilenet_v2_imagenet_int8.json b/...ntization/mobilenet_v2_imagenet_int8.json → ...ntization/mobilenet_v2_imagenet_int8.json
diff --git a/...obilenet_v2_imagenet_int8_per_tensor.json → ...obilenet_v2_imagenet_int8_per_tensor.json b/...obilenet_v2_imagenet_int8_per_tensor.json → ...obilenet_v2_imagenet_int8_per_tensor.json
diff --git a/...antization/mobilenetv3_imagenet_int8.json → ...antization/mobilenetv3_imagenet_int8.json b/...antization/mobilenetv3_imagenet_int8.json → ...antization/mobilenetv3_imagenet_int8.json
diff --git a/...nfigs/quantization/resnet50_imagenet.json → ...nfigs/quantization/resnet50_imagenet.json b/...nfigs/quantization/resnet50_imagenet.json → ...nfigs/quantization/resnet50_imagenet.json
diff --git a/.../quantization/resnet50_imagenet_int8.json → .../quantization/resnet50_imagenet_int8.json b/.../quantization/resnet50_imagenet_int8.json → .../quantization/resnet50_imagenet_int8.json
diff --git a/...on/resnet50_imagenet_int8_per_tensor.json → ...on/resnet50_imagenet_int8_per_tensor.json b/...on/resnet50_imagenet_int8_per_tensor.json → ...on/resnet50_imagenet_int8_per_tensor.json
diff --git a/...onfigs/quantization/rmnet_cifar_int8.json → ...onfigs/quantization/rmnet_cifar_int8.json b/...onfigs/quantization/rmnet_cifar_int8.json → ...onfigs/quantization/rmnet_cifar_int8.json
diff --git a/.../quantization/squeezenet1_1_imagenet.json → .../quantization/squeezenet1_1_imagenet.json b/.../quantization/squeezenet1_1_imagenet.json → .../quantization/squeezenet1_1_imagenet.json
diff --git a/...tization/squeezenet1_1_imagenet_int8.json → ...tization/squeezenet1_1_imagenet_int8.json b/...tization/squeezenet1_1_imagenet_int8.json → ...tization/squeezenet1_1_imagenet_int8.json
diff --git a/...ueezenet1_1_imagenet_int8_per_tensor.json → ...ueezenet1_1_imagenet_int8_per_tensor.json b/...ueezenet1_1_imagenet_int8_per_tensor.json → ...ueezenet1_1_imagenet_int8_per_tensor.json
diff --git a/...ty/inception_v3_imagenet_rb_sparsity.json → ...ty/inception_v3_imagenet_rb_sparsity.json b/...ty/inception_v3_imagenet_rb_sparsity.json → ...ty/inception_v3_imagenet_rb_sparsity.json
diff --git a/...mobilenet_v2_imagenet_const_sparsity.json → ...mobilenet_v2_imagenet_const_sparsity.json b/...mobilenet_v2_imagenet_const_sparsity.json → ...mobilenet_v2_imagenet_const_sparsity.json
diff --git a/...ty/mobilenet_v2_imagenet_rb_sparsity.json → ...ty/mobilenet_v2_imagenet_rb_sparsity.json b/...ty/mobilenet_v2_imagenet_rb_sparsity.json → ...ty/mobilenet_v2_imagenet_rb_sparsity.json
diff --git a/...ty/resnet34_cifar_magnitude_sparsity.json → ...ty/resnet34_cifar_magnitude_sparsity.json b/...ty/resnet34_cifar_magnitude_sparsity.json → ...ty/resnet34_cifar_magnitude_sparsity.json
diff --git a/.../sparsity/resnet34_cifar_rb_sparsity.json → .../sparsity/resnet34_cifar_rb_sparsity.json b/.../sparsity/resnet34_cifar_rb_sparsity.json → .../sparsity/resnet34_cifar_rb_sparsity.json
diff --git a/...arsity/resnet50_imagenet_rb_sparsity.json → ...arsity/resnet50_imagenet_rb_sparsity.json b/...arsity/resnet50_imagenet_rb_sparsity.json → ...arsity/resnet50_imagenet_rb_sparsity.json
diff --git a/...ception_v3_imagenet_rb_sparsity_int8.json → ...ception_v3_imagenet_rb_sparsity_int8.json b/...ception_v3_imagenet_rb_sparsity_int8.json → ...ception_v3_imagenet_rb_sparsity_int8.json
diff --git a/...bilenet_v2_imagenet_rb_sparsity_int8.json → ...bilenet_v2_imagenet_rb_sparsity_int8.json b/...bilenet_v2_imagenet_rb_sparsity_int8.json → ...bilenet_v2_imagenet_rb_sparsity_int8.json
diff --git a/...tion/resnet34_cifar_rb_sparsity_int8.json → ...tion/resnet34_cifar_rb_sparsity_int8.json b/...tion/resnet34_cifar_rb_sparsity_int8.json → ...tion/resnet34_cifar_rb_sparsity_int8.json
diff --git a/...resnet50_imagenet_rb_sparsity50_int8.json → ...resnet50_imagenet_rb_sparsity50_int8.json b/...resnet50_imagenet_rb_sparsity50_int8.json → ...resnet50_imagenet_rb_sparsity50_int8.json
diff --git a/...n/resnet50_imagenet_rb_sparsity_int8.json → ...n/resnet50_imagenet_rb_sparsity_int8.json b/...n/resnet50_imagenet_rb_sparsity_int8.json → ...n/resnet50_imagenet_rb_sparsity_int8.json
diff --git a/examples/classification/main.py → examples/torch/classification/main.py b/examples/classification/main.py → examples/torch/classification/main.py
@@ -36,23 +36,23 @@
 from torchvision.models import InceptionOutputs
 
 from nncf.common.utils.tensorboard import prepare_for_tensorboard
-from examples.common.argparser import get_common_argument_parser
-from examples.common.example_logger import logger
-from examples.common.execution import ExecutionMode, get_execution_mode, \
+from examples.torch.common.argparser import get_common_argument_parser
+from examples.torch.common.example_logger import logger
+from examples.torch.common.execution import ExecutionMode, get_execution_mode, \
     prepare_model_for_execution, start_worker
-from examples.common.model_loader import load_model
-from examples.common.optimizer import get_parameter_groups, make_optimizer
-from examples.common.sample_config import SampleConfig, create_sample_config
-from examples.common.utils import configure_logging, configure_paths, create_code_snapshot, \
+from examples.torch.common.model_loader import load_model
+from examples.torch.common.optimizer import get_parameter_groups, make_optimizer
+from examples.torch.common.sample_config import SampleConfig, create_sample_config
+from examples.torch.common.utils import configure_logging, configure_paths, create_code_snapshot, \
     print_args, make_additional_checkpoints, get_name, is_staged_quantization, \
     is_pretrained_model_requested, log_common_mlflow_params, SafeMLFLow, MockDataset, configure_device
-from examples.common.utils import write_metrics
+from examples.torch.common.utils import write_metrics
 from nncf import create_compressed_model
 from nncf.api.compression import CompressionStage
 from nncf.torch.dynamic_graph.graph_tracer import create_input_infos
 from nncf.torch.initialization import register_default_init_args, default_criterion_fn
 from nncf.torch.utils import safe_thread_call, is_main_process
-from examples.classification.common import set_seed, load_resuming_checkpoint
+from examples.torch.classification.common import set_seed, load_resuming_checkpoint
 
 model_names = sorted(name for name in models.__dict__
                      if name.islower() and not name.startswith("__")
@@ -99,7 +99,7 @@ def main(argv):
     if not is_staged_quantization(config):
         start_worker(main_worker, config)
     else:
-        from examples.classification.staged_quantization_worker import staged_quantization_main_worker
+        from examples.torch.classification.staged_quantization_worker import staged_quantization_main_worker
         start_worker(staged_quantization_main_worker, config)
 
 

diff --git a/examples/classification/models/__init__.py → ...s/torch/classification/models/__init__.py b/examples/classification/models/__init__.py → ...s/torch/classification/models/__init__.py
diff --git a/...assification/models/mobilenet_v2_32x32.py → ...assification/models/mobilenet_v2_32x32.py b/...assification/models/mobilenet_v2_32x32.py → ...assification/models/mobilenet_v2_32x32.py
diff --git a/...ssification/staged_quantization_worker.py → ...ssification/staged_quantization_worker.py b/...ssification/staged_quantization_worker.py → ...ssification/staged_quantization_worker.py
@@ -25,20 +25,20 @@
 from torchvision.models import InceptionOutputs
 
 from nncf.common.utils.tensorboard import prepare_for_tensorboard
-from examples.classification.main import create_data_loaders, validate, AverageMeter, accuracy, get_lr, \
+from examples.torch.classification.main import create_data_loaders, validate, AverageMeter, accuracy, get_lr, \
     create_datasets, inception_criterion_fn
-from examples.common.example_logger import logger
-from examples.common.execution import ExecutionMode, prepare_model_for_execution
-from examples.common.model_loader import load_model
-from examples.common.utils import configure_logging, print_args, make_additional_checkpoints, get_name, \
+from examples.torch.common.example_logger import logger
+from examples.torch.common.execution import ExecutionMode, prepare_model_for_execution
+from examples.torch.common.model_loader import load_model
+from examples.torch.common.utils import configure_logging, print_args, make_additional_checkpoints, get_name, \
     is_pretrained_model_requested, log_common_mlflow_params, SafeMLFLow, configure_device
 from nncf.torch.binarization.algo import BinarizationController
 from nncf.api.compression import CompressionStage
 from nncf.torch.initialization import register_default_init_args, default_criterion_fn
 from nncf.torch.model_creation import create_compressed_model
 from nncf.torch.quantization.algo import QuantizationController
 from nncf.torch.utils import is_main_process
-from examples.classification.common import set_seed, load_resuming_checkpoint
+from examples.torch.classification.common import set_seed, load_resuming_checkpoint
 
 
 class KDLossCalculator:

diff --git a/examples/common/__init__.py → examples/torch/common/__init__.py b/examples/common/__init__.py → examples/torch/common/__init__.py
diff --git a/examples/common/argparser.py → examples/torch/common/argparser.py b/examples/common/argparser.py → examples/torch/common/argparser.py
@@ -11,7 +11,7 @@
  limitations under the License.
 """
 
-from examples.common.sample_config import CustomArgumentParser
+from examples.torch.common.sample_config import CustomArgumentParser
 from nncf.common.hardware.config import HWConfigType
 
 

diff --git a/examples/common/distributed.py → examples/torch/common/distributed.py b/examples/common/distributed.py → examples/torch/common/distributed.py
@@ -18,7 +18,7 @@
 from torch import distributed as dist
 from torch.utils.data import Sampler
 
-from examples.common.example_logger import logger
+from examples.torch.common.example_logger import logger
 
 
 def configure_distributed(config):

diff --git a/examples/common/example_logger.py → examples/torch/common/example_logger.py b/examples/common/example_logger.py → examples/torch/common/example_logger.py
diff --git a/examples/common/execution.py → examples/torch/common/execution.py b/examples/common/execution.py → examples/torch/common/execution.py
@@ -13,7 +13,7 @@
 
 import torch
 import torch.multiprocessing as mp
-from examples.common.sample_config import SampleConfig
+from examples.torch.common.sample_config import SampleConfig
 
 
 class ExecutionMode:

diff --git a/examples/common/model_loader.py → examples/torch/common/model_loader.py b/examples/common/model_loader.py → examples/torch/common/model_loader.py
@@ -16,12 +16,12 @@
 import torchvision.models
 from functools import partial
 
-import examples.common.models as custom_models
-from examples.common.example_logger import logger
-import examples.common.restricted_pickle_module as restricted_pickle_module
+import examples.torch.common.models as custom_models
+from examples.torch.common.example_logger import logger
+import examples.torch.common.restricted_pickle_module as restricted_pickle_module
 from nncf import load_state
 from nncf.torch.utils import safe_thread_call
-from examples.classification.models.mobilenet_v2_32x32 import MobileNetV2For32x32
+from examples.torch.classification.models.mobilenet_v2_32x32 import MobileNetV2For32x32
 
 
 def load_model(model, pretrained=True, num_classes=1000, model_params=None,

diff --git a/examples/torch/common/models/__init__.py b/examples/torch/common/models/__init__.py
@@ -0,0 +1,2 @@
+from examples.torch.common.models.segmentation import *
+from examples.torch.common.models.classification import *
diff --git a/.../common/models/classification/__init__.py → .../common/models/classification/__init__.py b/.../common/models/classification/__init__.py → .../common/models/classification/__init__.py
diff --git a/...ls/classification/inceptionv3_cifar100.py → ...ls/classification/inceptionv3_cifar100.py b/...ls/classification/inceptionv3_cifar100.py → ...ls/classification/inceptionv3_cifar100.py
diff --git a/...mmon/models/classification/mobilenetv3.py → ...mmon/models/classification/mobilenetv3.py b/...mmon/models/classification/mobilenetv3.py → ...mmon/models/classification/mobilenetv3.py
diff --git a/...mon/models/classification/resnet_cifar.py → ...mon/models/classification/resnet_cifar.py b/...mon/models/classification/resnet_cifar.py → ...mon/models/classification/resnet_cifar.py
diff --git a/...mmon/models/classification/rmnet_cifar.py → ...mmon/models/classification/rmnet_cifar.py b/...mmon/models/classification/rmnet_cifar.py → ...mmon/models/classification/rmnet_cifar.py
diff --git a/...es/common/models/segmentation/__init__.py → ...ch/common/models/segmentation/__init__.py b/...es/common/models/segmentation/__init__.py → ...ch/common/models/segmentation/__init__.py
diff --git a/examples/common/models/segmentation/enet.py → .../torch/common/models/segmentation/enet.py b/examples/common/models/segmentation/enet.py → .../torch/common/models/segmentation/enet.py
@@ -17,7 +17,7 @@
 import torch.nn as nn
 import torch
 
-from examples.common.example_logger import logger
+from examples.torch.common.example_logger import logger
 
 class InitialBlock(nn.Module):
     """The initial block is composed of two branches:

diff --git a/examples/common/models/segmentation/icnet.py → ...torch/common/models/segmentation/icnet.py b/examples/common/models/segmentation/icnet.py → ...torch/common/models/segmentation/icnet.py
@@ -26,7 +26,7 @@
 import torch
 import torch.nn.functional as F
 
-from examples.common.example_logger import logger
+from examples.torch.common.example_logger import logger
 from nncf.torch.utils import is_tracing_state
 
 

diff --git a/examples/common/models/segmentation/unet.py → .../torch/common/models/segmentation/unet.py b/examples/common/models/segmentation/unet.py → .../torch/common/models/segmentation/unet.py
@@ -22,7 +22,7 @@
 
 from nncf.torch.utils import is_tracing_state
 
-from examples.common.example_logger import logger
+from examples.torch.common.example_logger import logger
 
 class UNet(nn.Module):
     def __init__(

diff --git a/examples/common/optimizer.py → examples/torch/common/optimizer.py b/examples/common/optimizer.py → examples/torch/common/optimizer.py
diff --git a/examples/common/restricted_pickle_module.py → .../torch/common/restricted_pickle_module.py b/examples/common/restricted_pickle_module.py → .../torch/common/restricted_pickle_module.py
diff --git a/examples/common/sample_config.py → examples/torch/common/sample_config.py b/examples/common/sample_config.py → examples/torch/common/sample_config.py
diff --git a/examples/common/utils.py → examples/torch/common/utils.py b/examples/common/utils.py → examples/torch/common/utils.py
@@ -27,14 +27,14 @@
 from PIL import Image
 import torch.utils.data as data
 
-from examples.common.distributed import configure_distributed
-from examples.common.execution import ExecutionMode, get_device
-from examples.common.sample_config import SampleConfig
+from examples.torch.common.distributed import configure_distributed
+from examples.torch.common.execution import ExecutionMode, get_device
+from examples.torch.common.sample_config import SampleConfig
 from torch.utils.tensorboard import SummaryWriter
 import mlflow
 import torch
 
-from examples.common.example_logger import logger as default_logger
+from examples.torch.common.example_logger import logger as default_logger
 from nncf.torch.utils import is_main_process
 
 # pylint: disable=import-error

diff --git a/examples/object_detection/README.md → examples/torch/object_detection/README.md b/examples/object_detection/README.md → examples/torch/object_detection/README.md
@@ -14,7 +14,7 @@ This sample demonstrates DL model compression capabailites for object detection
 To work with the sample you should install the corresponding Python package dependencies
 
 ```
-pip install -r examples/requirements.txt
+pip install -r examples/torch/requirements.txt
 ```
 
 ## Quantize FP32 pretrained model
@@ -25,7 +25,7 @@ This scenario demonstrates quantization with fine-tuning of SSD300 on VOC datase
 
 #### Run object detection sample
 - If you did not install the package then add the repository root folder to the `PYTHONPATH` environment variable
-- Navigate to the `examples/object_detection` folder
+- Navigate to the `examples/torch/object_detection` folder
 - Run the following command to start compression with fine-tuning on GPUs:
 `python main.py -m train --config configs/ssd300_vgg_int8_voc.json --data <path_to_dataset> --log-dir=../../results/quantization/ssd300_int8`
 It may take a few epochs to get the baseline accuracy results.

diff --git a/examples/object_detection/__init__.py → examples/torch/object_detection/__init__.py b/examples/object_detection/__init__.py → examples/torch/object_detection/__init__.py
diff --git a/...tection/configs/ssd300_mobilenet_voc.json → ...tection/configs/ssd300_mobilenet_voc.json b/...tection/configs/ssd300_mobilenet_voc.json → ...tection/configs/ssd300_mobilenet_voc.json
diff --git a/.../ssd300_mobilenet_voc_magnitude_int8.json → .../ssd300_mobilenet_voc_magnitude_int8.json b/.../ssd300_mobilenet_voc_magnitude_int8.json → .../ssd300_mobilenet_voc_magnitude_int8.json
diff --git a/...sd300_mobilenet_voc_rb_sparsity_int8.json → ...sd300_mobilenet_voc_rb_sparsity_int8.json b/...sd300_mobilenet_voc_rb_sparsity_int8.json → ...sd300_mobilenet_voc_rb_sparsity_int8.json
diff --git a/...ect_detection/configs/ssd300_vgg_voc.json → ...ect_detection/configs/ssd300_vgg_voc.json b/...ect_detection/configs/ssd300_vgg_voc.json → ...ect_detection/configs/ssd300_vgg_voc.json
diff --git a/...etection/configs/ssd300_vgg_voc_int8.json → ...etection/configs/ssd300_vgg_voc_int8.json b/...etection/configs/ssd300_vgg_voc_int8.json → ...etection/configs/ssd300_vgg_voc_int8.json
diff --git a/...d300_vgg_voc_magnitude_sparsity_int8.json → ...d300_vgg_voc_magnitude_sparsity_int8.json b/...d300_vgg_voc_magnitude_sparsity_int8.json → ...d300_vgg_voc_magnitude_sparsity_int8.json
diff --git a/...300_vgg_voc_pruning_geometric_median.json → ...300_vgg_voc_pruning_geometric_median.json b/...300_vgg_voc_pruning_geometric_median.json → ...300_vgg_voc_pruning_geometric_median.json
diff --git a/...ect_detection/configs/ssd512_vgg_voc.json → ...ect_detection/configs/ssd512_vgg_voc.json b/...ect_detection/configs/ssd512_vgg_voc.json → ...ect_detection/configs/ssd512_vgg_voc.json
diff --git a/...etection/configs/ssd512_vgg_voc_int8.json → ...etection/configs/ssd512_vgg_voc_int8.json b/...etection/configs/ssd512_vgg_voc_int8.json → ...etection/configs/ssd512_vgg_voc_int8.json
diff --git a/...d512_vgg_voc_magnitude_sparsity_int8.json → ...d512_vgg_voc_magnitude_sparsity_int8.json b/...d512_vgg_voc_magnitude_sparsity_int8.json → ...d512_vgg_voc_magnitude_sparsity_int8.json
diff --git a/examples/object_detection/dataset.py → examples/torch/object_detection/dataset.py b/examples/object_detection/dataset.py → examples/torch/object_detection/dataset.py
@@ -16,9 +16,9 @@
 import cv2
 import numpy as np
 import torch
-from examples.object_detection.datasets.coco import COCODataset
-from examples.object_detection.datasets.voc0712 import VOCDetection, VOCAnnotationTransform
-from examples.object_detection.utils.augmentations import SSDAugmentation
+from examples.torch.object_detection.datasets.coco import COCODataset
+from examples.torch.object_detection.datasets.voc0712 import VOCDetection, VOCAnnotationTransform
+from examples.torch.object_detection.utils.augmentations import SSDAugmentation
 from nncf.torch.dynamic_graph.graph_tracer import create_input_infos
 
 VOC_MEAN = (0.406, 0.456, 0.485)

diff --git a/...les/object_detection/datasets/__init__.py → ...rch/object_detection/datasets/__init__.py b/...les/object_detection/datasets/__init__.py → ...rch/object_detection/datasets/__init__.py
diff --git a/examples/object_detection/datasets/coco.py → ...s/torch/object_detection/datasets/coco.py b/examples/object_detection/datasets/coco.py → ...s/torch/object_detection/datasets/coco.py
@@ -21,7 +21,7 @@
 import torch
 import torch.utils.data as data
 
-from examples.common.example_logger import logger
+from examples.torch.common.example_logger import logger
 
 COCO_CLASSES = (  # always index 0
     "person", "bicycle", "car", "motorcycle", "airplane", "bus", "train", "truck", "boat", "traffic light",

diff --git a/...ples/object_detection/datasets/voc0712.py → ...orch/object_detection/datasets/voc0712.py b/...ples/object_detection/datasets/voc0712.py → ...orch/object_detection/datasets/voc0712.py
diff --git a/examples/object_detection/eval.py → examples/torch/object_detection/eval.py b/examples/object_detection/eval.py → examples/torch/object_detection/eval.py
@@ -23,7 +23,7 @@
 from torch import distributed as dist
 from torch.nn import functional as F
 
-from examples.common.example_logger import logger
+from examples.torch.common.example_logger import logger
 from nncf.torch.utils import is_main_process, is_dist_avail_and_initialized, get_world_size
 
 

diff --git a/examples/object_detection/layers/__init__.py → ...torch/object_detection/layers/__init__.py b/examples/object_detection/layers/__init__.py → ...torch/object_detection/layers/__init__.py
diff --git a/...ples/object_detection/layers/box_utils.py → ...orch/object_detection/layers/box_utils.py b/...ples/object_detection/layers/box_utils.py → ...orch/object_detection/layers/box_utils.py
diff --git a/...t_detection/layers/extensions/__init__.py → ...t_detection/layers/extensions/__init__.py b/...t_detection/layers/extensions/__init__.py → ...t_detection/layers/extensions/__init__.py
diff --git a/...etection/layers/extensions/extensions.cpp → ...etection/layers/extensions/extensions.cpp b/...etection/layers/extensions/extensions.cpp → ...etection/layers/extensions/extensions.cpp
diff --git a/...t_detection/layers/extensions/nms/nms.cpp → ...t_detection/layers/extensions/nms/nms.cpp b/...t_detection/layers/extensions/nms/nms.cpp → ...t_detection/layers/extensions/nms/nms.cpp
diff --git a/...ect_detection/layers/extensions/nms/nms.h → ...ect_detection/layers/extensions/nms/nms.h b/...ect_detection/layers/extensions/nms/nms.h → ...ect_detection/layers/extensions/nms/nms.h
diff --git a/...ction/layers/extensions/nms/nms_kernel.cu → ...ction/layers/extensions/nms/nms_kernel.cu b/...ction/layers/extensions/nms/nms_kernel.cu → ...ction/layers/extensions/nms/nms_kernel.cu
diff --git a/...ct_detection/layers/functions/__init__.py → ...ct_detection/layers/functions/__init__.py b/...ct_detection/layers/functions/__init__.py → ...ct_detection/layers/functions/__init__.py
diff --git a/...t_detection/layers/functions/detection.py → ...t_detection/layers/functions/detection.py b/...t_detection/layers/functions/detection.py → ...t_detection/layers/functions/detection.py
diff --git a/...t_detection/layers/functions/prior_box.py → ...t_detection/layers/functions/prior_box.py b/...t_detection/layers/functions/prior_box.py → ...t_detection/layers/functions/prior_box.py
diff --git a/...ject_detection/layers/modules/__init__.py → ...ject_detection/layers/modules/__init__.py b/...ject_detection/layers/modules/__init__.py → ...ject_detection/layers/modules/__init__.py
diff --git a/...object_detection/layers/modules/l2norm.py → ...object_detection/layers/modules/l2norm.py b/...object_detection/layers/modules/l2norm.py → ...object_detection/layers/modules/l2norm.py
diff --git a/...detection/layers/modules/multibox_loss.py → ...detection/layers/modules/multibox_loss.py b/...detection/layers/modules/multibox_loss.py → ...detection/layers/modules/multibox_loss.py
diff --git a/...ject_detection/layers/modules/ssd_head.py → ...ject_detection/layers/modules/ssd_head.py b/...ject_detection/layers/modules/ssd_head.py → ...ject_detection/layers/modules/ssd_head.py
@@ -15,7 +15,7 @@
 from torch import nn
 from torch.nn import functional as F
 
-from examples.object_detection.layers import DetectionOutput, PriorBox
+from examples.torch.object_detection.layers import DetectionOutput, PriorBox
 from nncf.torch.dynamic_graph.context import no_nncf_trace
 
 

diff --git a/examples/object_detection/main.py → examples/torch/object_detection/main.py b/examples/object_detection/main.py → examples/torch/object_detection/main.py
@@ -20,27 +20,27 @@
 import torch
 import torch.utils.data as data
 
-from examples.common import restricted_pickle_module
-from examples.common.model_loader import load_resuming_model_state_dict_and_checkpoint_from_path
-from examples.common.sample_config import create_sample_config, SampleConfig
+from examples.torch.common import restricted_pickle_module
+from examples.torch.common.model_loader import load_resuming_model_state_dict_and_checkpoint_from_path
+from examples.torch.common.sample_config import create_sample_config, SampleConfig
 from torch.optim.lr_scheduler import ReduceLROnPlateau
 
-from examples.common.argparser import get_common_argument_parser
-from examples.common.distributed import DistributedSampler
-from examples.common.example_logger import logger
-from examples.common.execution import get_execution_mode
-from examples.common.execution import prepare_model_for_execution, start_worker
+from examples.torch.common.argparser import get_common_argument_parser
+from examples.torch.common.distributed import DistributedSampler
+from examples.torch.common.example_logger import logger
+from examples.torch.common.execution import get_execution_mode
+from examples.torch.common.execution import prepare_model_for_execution, start_worker
 from nncf.api.compression import CompressionStage
 from nncf.torch.initialization import register_default_init_args
-from examples.common.optimizer import get_parameter_groups, make_optimizer
-from examples.common.utils import get_name, make_additional_checkpoints, configure_paths, \
+from examples.torch.common.optimizer import get_parameter_groups, make_optimizer
+from examples.torch.common.utils import get_name, make_additional_checkpoints, configure_paths, \
     create_code_snapshot, is_on_first_rank, configure_logging, print_args, is_pretrained_model_requested, \
     log_common_mlflow_params, SafeMLFLow, configure_device
-from examples.common.utils import write_metrics
-from examples.object_detection.dataset import detection_collate, get_testing_dataset, get_training_dataset
-from examples.object_detection.eval import test_net
-from examples.object_detection.layers.modules import MultiBoxLoss
-from examples.object_detection.model import build_ssd
+from examples.torch.common.utils import write_metrics
+from examples.torch.object_detection.dataset import detection_collate, get_testing_dataset, get_training_dataset
+from examples.torch.object_detection.eval import test_net
+from examples.torch.object_detection.layers.modules import MultiBoxLoss
+from examples.torch.object_detection.model import build_ssd
 from nncf import create_compressed_model, load_state
 from nncf.torch.dynamic_graph.graph_tracer import create_input_infos
 from nncf.torch.utils import is_main_process

diff --git a/examples/object_detection/model.py → examples/torch/object_detection/model.py b/examples/object_detection/model.py → examples/torch/object_detection/model.py
@@ -11,8 +11,8 @@
  limitations under the License.
 """
 
-from examples.object_detection.models.ssd_mobilenet import build_ssd_mobilenet
-from examples.object_detection.models.ssd_vgg import build_ssd_vgg
+from examples.torch.object_detection.models.ssd_mobilenet import build_ssd_mobilenet
+from examples.torch.object_detection.models.ssd_vgg import build_ssd_vgg
 
 
 def build_ssd(net_name, cfg, ssd_dim, num_classes, config):

diff --git a/examples/object_detection/models/__init__.py → ...torch/object_detection/models/__init__.py b/examples/object_detection/models/__init__.py → ...torch/object_detection/models/__init__.py
diff --git a/.../object_detection/models/ssd_mobilenet.py → .../object_detection/models/ssd_mobilenet.py b/.../object_detection/models/ssd_mobilenet.py → .../object_detection/models/ssd_mobilenet.py
@@ -14,9 +14,9 @@
 import torch
 import torch.nn as nn
 
-from examples.common import restricted_pickle_module
-from examples.common.example_logger import logger
-from examples.object_detection.layers.modules.ssd_head import MultiOutputSequential, SSDDetectionOutput
+from examples.torch.common import restricted_pickle_module
+from examples.torch.common.example_logger import logger
+from examples.torch.object_detection.layers.modules.ssd_head import MultiOutputSequential, SSDDetectionOutput
 from nncf.torch.checkpoint_loading import load_state
 
 

diff --git a/examples/object_detection/models/ssd_vgg.py → .../torch/object_detection/models/ssd_vgg.py b/examples/object_detection/models/ssd_vgg.py → .../torch/object_detection/models/ssd_vgg.py
@@ -15,10 +15,10 @@
 import torch
 import torch.nn as nn
 
-from examples.common import restricted_pickle_module
-from examples.common.example_logger import logger
-from examples.object_detection.layers import L2Norm
-from examples.object_detection.layers.modules.ssd_head import MultiOutputSequential, SSDDetectionOutput
+from examples.torch.common import restricted_pickle_module
+from examples.torch.common.example_logger import logger
+from examples.torch.object_detection.layers import L2Norm
+from examples.torch.object_detection.layers.modules.ssd_head import MultiOutputSequential, SSDDetectionOutput
 from nncf.torch.checkpoint_loading import load_state
 
 BASE_NUM_OUTPUTS = {

diff --git a/examples/object_detection/utils/__init__.py → .../torch/object_detection/utils/__init__.py b/examples/object_detection/utils/__init__.py → .../torch/object_detection/utils/__init__.py
diff --git a/...s/object_detection/utils/augmentations.py → ...h/object_detection/utils/augmentations.py b/...s/object_detection/utils/augmentations.py → ...h/object_detection/utils/augmentations.py
diff --git a/examples/requirements.txt → examples/torch/requirements.txt b/examples/requirements.txt → examples/torch/requirements.txt
diff --git a/examples/semantic_segmentation/README.md → ...les/torch/semantic_segmentation/README.md b/examples/semantic_segmentation/README.md → ...les/torch/semantic_segmentation/README.md
@@ -14,7 +14,7 @@ This sample demonstrates DL model compression capabilities for semantic segmenta
 To work with the sample you should install the corresponding Python package dependencies
 
 ```
-pip install -r examples/requirements.txt
+pip install -r examples/torch/requirements.txt
 ```
 
 ## Quantize FP32 pretrained model
@@ -25,7 +25,7 @@ This scenario demonstrates quantization with fine-tuning of UNet on Mapillary Vi
 
 #### Run semantic segmentation sample
 - If you did not install the package then add the repository root folder to the `PYTHONPATH` environment variable
-- Navigate to the `examples/segmentation` folder
+- Navigate to the `examples/torch/segmentation` folder
 - Run the following command to start compression with fine-tuning on GPUs:
 `python main.py -m train --config configs/unet_mapillary_int8.json --data <path_to_dataset> --weights <path_to_fp32_model_checkpoint>`
 

diff --git a/...ic_segmentation/configs/icnet_camvid.json → ...ic_segmentation/configs/icnet_camvid.json b/...ic_segmentation/configs/icnet_camvid.json → ...ic_segmentation/configs/icnet_camvid.json
diff --git a/...gmentation/configs/icnet_camvid_int8.json → ...gmentation/configs/icnet_camvid_int8.json b/...gmentation/configs/icnet_camvid_int8.json → ...gmentation/configs/icnet_camvid_int8.json
diff --git a/...icnet_camvid_magnitude_sparsity_int8.json → ...icnet_camvid_magnitude_sparsity_int8.json b/...icnet_camvid_magnitude_sparsity_int8.json → ...icnet_camvid_magnitude_sparsity_int8.json
diff --git a/...egmentation/configs/icnet_cityscapes.json → ...egmentation/configs/icnet_cityscapes.json b/...egmentation/configs/icnet_cityscapes.json → ...egmentation/configs/icnet_cityscapes.json
diff --git a/...tic_segmentation/configs/unet_camvid.json → ...tic_segmentation/configs/unet_camvid.json b/...tic_segmentation/configs/unet_camvid.json → ...tic_segmentation/configs/unet_camvid.json
diff --git a/...ntation/configs/unet_camvid_bin_xnor.json → ...ntation/configs/unet_camvid_bin_xnor.json b/...ntation/configs/unet_camvid_bin_xnor.json → ...ntation/configs/unet_camvid_bin_xnor.json
diff --git a/...egmentation/configs/unet_camvid_int8.json → ...egmentation/configs/unet_camvid_int8.json b/...egmentation/configs/unet_camvid_int8.json → ...egmentation/configs/unet_camvid_int8.json
diff --git a/.../unet_camvid_magnitude_sparsity_int8.json → .../unet_camvid_magnitude_sparsity_int8.json b/.../unet_camvid_magnitude_sparsity_int8.json → .../unet_camvid_magnitude_sparsity_int8.json
diff --git a/...tion/configs/unet_camvid_rb_sparsity.json → ...tion/configs/unet_camvid_rb_sparsity.json b/...tion/configs/unet_camvid_rb_sparsity.json → ...tion/configs/unet_camvid_rb_sparsity.json
diff --git a/...segmentation/configs/unet_cityscapes.json → ...segmentation/configs/unet_cityscapes.json b/...segmentation/configs/unet_cityscapes.json → ...segmentation/configs/unet_cityscapes.json
diff --git a/..._segmentation/configs/unet_mapillary.json → ..._segmentation/configs/unet_mapillary.json b/..._segmentation/configs/unet_mapillary.json → ..._segmentation/configs/unet_mapillary.json
diff --git a/...entation/configs/unet_mapillary_int8.json → ...entation/configs/unet_mapillary_int8.json b/...entation/configs/unet_mapillary_int8.json → ...entation/configs/unet_mapillary_int8.json
diff --git a/...gs/unet_mapillary_magnitude_sparsity.json → ...gs/unet_mapillary_magnitude_sparsity.json b/...gs/unet_mapillary_magnitude_sparsity.json → ...gs/unet_mapillary_magnitude_sparsity.json
diff --git a/...et_mapillary_magnitude_sparsity_int8.json → ...et_mapillary_magnitude_sparsity_int8.json b/...et_mapillary_magnitude_sparsity_int8.json → ...et_mapillary_magnitude_sparsity_int8.json
diff --git a/...t_mapillary_pruning_geometric_median.json → ...t_mapillary_pruning_geometric_median.json b/...t_mapillary_pruning_geometric_median.json → ...t_mapillary_pruning_geometric_median.json
diff --git a/...mantic_segmentation/configs/unet_voc.json → ...mantic_segmentation/configs/unet_voc.json b/...mantic_segmentation/configs/unet_voc.json → ...mantic_segmentation/configs/unet_voc.json
diff --git a/...emantic_segmentation/datasets/__init__.py → ...emantic_segmentation/datasets/__init__.py b/...emantic_segmentation/datasets/__init__.py → ...emantic_segmentation/datasets/__init__.py
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1,2 @@
		from examples.torch.common.models.segmentation import *
		from examples.torch.common.models.classification import *