refactor packages to avoid confusion with environment packages with t…

…he same name (#757)

* refactor packages to avoid confusion with environment packages with the same name

---------

Co-authored-by: elad cohen <elad.cohen@altair-semi.com>

tutorials/quick_start/README.md

@@ -60,12 +60,12 @@ The weights modification is done through a knowledge distillation technique sour
 
 To execute the model compression with this approach, run the following Python script:
 ```python
-python main.py --model_name mobilenet_v2 --model_library torchvision --gptq --validation_dataset_folder <my path> --representative_dataset_folder <my path> 
+python main.py --model_name mobilenet_v2 --model_library torchvision_lib --gptq --validation_dataset_folder <my path> --representative_dataset_folder <my path> 
 ```
 
 Please note that the Mixed-Precision and Gradient-based Post Training Quantization (GPTQ) strategies can be combined to achieve a more significant model compression while mitigating the impact on model performance.
 #### More examples
-More details and examples for using Ultrlytics models can be found in this [readme](./pytorch_fw/ultralytics/README.md)   
+More details and examples for using Ultrlytics models can be found in this [readme](./pytorch_fw/ultralytics_lib/README.md)   
 
 ## Results
 The latest performance results of MCT on various of models can be found in the [results CSV](./results/model_quantization_results.csv) table. 
@@ -83,7 +83,7 @@ The following external packages were tested with this project:
 This project is licensed under [Apache License 2.0](../../LICENSE.md).
 However, please note that different external packages have their own licenses. When using this project, you have the option to choose one of the following external packages:
 
-- torchvision: [license link](https://github.com/UiPath/torchvision/blob/master/LICENSE), [license copy](./pytorch_fw/torchvision/LICENSE)
-- timm: [license link](https://github.com/huggingface/pytorch-image-models/blob/main/LICENSE), [license copy](./pytorch_fw/timm/LICENSE)
-- ultralytics: [license link](https://github.com/ultralytics/ultralytics/blob/main/LICENSE), [license copy](./pytorch_fw/ultralytics/LICENSE)
+- torchvision: [license link](https://github.com/UiPath/torchvision/blob/master/LICENSE), [license copy](./pytorch_fw/torchvision_lib/LICENSE)
+- timm: [license link](https://github.com/huggingface/pytorch-image-models/blob/main/LICENSE), [license copy](./pytorch_fw/timm_lib/LICENSE)
+- ultralytics: [license link](https://github.com/ultralytics/ultralytics/blob/main/LICENSE), [license copy](./pytorch_fw/ultralytics_lib/LICENSE)
 - keras-applications: [license link](https://github.com/keras-team/keras-applications/blob/master/LICENSE), [license copy](./keras_fw/keras_applications/LICENSE)

tutorials/quick_start/common/utils.py

@@ -33,12 +33,12 @@ def find_modules(lib: str) -> Tuple[str, str]:
 
     # Search in PyTorch libraries
     if importlib.util.find_spec('pytorch_fw.' + lib) is not None:
-        model_lib_module = 'pytorch_fw.' + lib + '.model_lib_' + lib
+        model_lib_module = 'pytorch_fw.' + lib + '.model_lib'
         quant_module = 'pytorch_fw.quant'
 
     # Search in Keras libraries
     elif importlib.util.find_spec('keras_fw.' + lib) is not None:
-        model_lib_module = 'keras_fw.' + lib + '.model_lib_' + lib
+        model_lib_module = 'keras_fw.' + lib + '.model_lib'
         quant_module = 'keras_fw.quant'
     else:
         raise Exception(f'Error: model library {lib} is not supported')

tutorials/quick_start/pytorch_fw/timm/model_lib_timm.py → tutorials/quick_start/pytorch_fw/timm_lib/model_lib.py

@@ -15,11 +15,11 @@
 import timm
 from timm.data import create_dataset, create_loader, resolve_data_config
 
-from common.model_lib import BaseModelLib
-from pytorch_fw.utils import classification_eval, get_representative_dataset
-from common.constants import MODEL_NAME, BATCH_SIZE, VALIDATION_SET_LIMIT, VALIDATION_DATASET_FOLDER, \
-    IMAGENET_DATASET
-from common.results import DatasetInfo
+from tutorials.quick_start.common.model_lib import BaseModelLib
+from tutorials.quick_start.pytorch_fw.utils import classification_eval, get_representative_dataset
+from tutorials.quick_start.common.constants import MODEL_NAME, BATCH_SIZE, VALIDATION_SET_LIMIT, \
+    VALIDATION_DATASET_FOLDER, IMAGENET_DATASET
+from tutorials.quick_start.common.results import DatasetInfo
 
 
 class ModelLib(BaseModelLib):

tutorials/quick_start/pytorch_fw/torchvision/model_lib_torchvision.py → tutorials/quick_start/pytorch_fw/torchvision_lib/model_lib.py

@@ -17,11 +17,12 @@
 from torchvision.models import list_models, get_model, get_model_weights, get_weight
 from torch.utils.data import Subset
 
-from common.model_lib import BaseModelLib
-from pytorch_fw.utils import classification_eval, get_representative_dataset
-from common.constants import MODEL_NAME, BATCH_SIZE, VALIDATION_SET_LIMIT, VALIDATION_DATASET_FOLDER, IMAGENET_DATASET
+from tutorials.quick_start.common.model_lib import BaseModelLib
+from tutorials.quick_start.pytorch_fw.utils import classification_eval, get_representative_dataset
+from tutorials.quick_start.common.constants import MODEL_NAME, BATCH_SIZE, VALIDATION_SET_LIMIT, \
+    VALIDATION_DATASET_FOLDER, IMAGENET_DATASET
 
-from common.results import DatasetInfo
+from tutorials.quick_start.common.results import DatasetInfo
 
 
 class ModelLib(BaseModelLib):

tutorials/quick_start/pytorch_fw/ultralytics/README.md → tutorials/quick_start/pytorch_fw/ultralytics_lib/README.md

@@ -22,11 +22,11 @@ setenv PYTHONPATH /path/to/model_optimization/folder
 ```
 
 ```csh 
-python main.py --model_name yolov8n --model_library ultralytics --representative_dataset_folder <download_location>/datasets/coco/images/train2017 --num_representative_images 64 --batch_size 1 
+python main.py --model_name yolov8n --model_library ultralytics_lib --representative_dataset_folder <download_location>/datasets/coco/images/train2017 --num_representative_images 64 --batch_size 1 
 ```
 In this example, we are running the main.py script with the following parameters:
 - `--model_name`: Specifies the name of the model to be used out of Ultralytics models (yolov8n in this case).
-- `--model_library`: Specifies the package of the pre-trained models from which the model name is taken (in this case, ultralytics).
+- `--model_library`: Specifies the package of the pre-trained models from which the model name is taken (in this case, ultralytics_lib).
 - `--representative_dataset_folder`: Specifies the path to the local copy of the dataset to be used for quantization. In this case, we use the 'train' split of the downloaded dataset provided by Ultralytics.
 - `--num_representative_images`: Specifies the number of representative images to be used for quantization.
 - `--batch_size`: Specifies the batch size to be used.

tutorials/quick_start/results/model_quantization_results.csv

@@ -1,13 +1,13 @@
 ModelName,ModelLibrary,DatasetName,TotalImages,FloatAcc,QuantAcc,Size[MB],BitsConfig,QuantTechnique,TPC
-mobilenet_v2,torchvision,ImageNet,50000,0.7215,0.7188,3.47,W8A8,PTQ,default_pytorch_tpc-v4
-regnet_y_400mf,torchvision,ImageNet,50000,0.7578,0.7542,4.31,W8A8,PTQ,default_pytorch_tpc-v4
-shufflenet_v2_x0_5,torchvision,ImageNet,50000,0.6055,0.597,1.36,W8A8,PTQ,default_pytorch_tpc-v4
-squeezenet1_0,torchvision,ImageNet,50000,0.581,0.5767,1.24,W8A8,PTQ,default_pytorch_tpc-v4
-regnetx_002,timm,ImageNet,50000,0.6876,0.6827,2.67,W8A8,PTQ,default_pytorch_tpc-v4
-regnety_008,timm,ImageNet,50000,0.7632,0.7598,6.23,W8A8,PTQ,default_pytorch_tpc-v4
-resnet10t,timm,ImageNet,50000,0.6656,0.6643,5.43,W8A8,PTQ,default_pytorch_tpc-v4
-resnet18,timm,ImageNet,50000,0.6976,0.6961,11.68,W8A8,PTQ,default_pytorch_tpc-v4
-efficientnet_es,timm,ImageNet,50000,0.7808,0.7774,5.39,W8A8,PTQ,default_pytorch_tpc-v4
+mobilenet_v2,torchvision_lib,ImageNet,50000,0.7215,0.7188,3.47,W8A8,PTQ,default_pytorch_tpc-v4
+regnet_y_400mf,torchvision_lib,ImageNet,50000,0.7578,0.7542,4.31,W8A8,PTQ,default_pytorch_tpc-v4
+shufflenet_v2_x0_5,torchvision_lib,ImageNet,50000,0.6055,0.597,1.36,W8A8,PTQ,default_pytorch_tpc-v4
+squeezenet1_0,torchvision_lib,ImageNet,50000,0.581,0.5767,1.24,W8A8,PTQ,default_pytorch_tpc-v4
+regnetx_002,timm_lib,ImageNet,50000,0.6876,0.6827,2.67,W8A8,PTQ,default_pytorch_tpc-v4
+regnety_008,timm_lib,ImageNet,50000,0.7632,0.7598,6.23,W8A8,PTQ,default_pytorch_tpc-v4
+resnet10t,timm_lib,ImageNet,50000,0.6656,0.6643,5.43,W8A8,PTQ,default_pytorch_tpc-v4
+resnet18,timm_lib,ImageNet,50000,0.6976,0.6961,11.68,W8A8,PTQ,default_pytorch_tpc-v4
+efficientnet_es,timm_lib,ImageNet,50000,0.7808,0.7774,5.39,W8A8,PTQ,default_pytorch_tpc-v4
 yolov8n,ultralytics,COCO,5000,0.3704,0.3617,3.15,W8A8,PTQ,default_pytorch_tpc-v4
 yolov8m,ultralytics,COCO,5000,0.4999,0.494,25.9,W8A8,PTQ,default_pytorch_tpc-v4
 mobilenet_v2.MobileNetV2,keras_applications,ImageNet,50000,0.7185,0.7157,3.47,W8A8,PTQ,default_keras_tpc-v4