diff --git a/.github/workflows/base.yml.template b/.github/workflows/base.yml.template
index 27c53d186..68e89aa4e 100644
--- a/.github/workflows/base.yml.template
+++ b/.github/workflows/base.yml.template
@@ -57,6 +57,11 @@ jobs:
         mkdir $CONDA_PKGS_DIRS
         mkdir $PYTORCH_PKG_CACHE_PATH
 
+    - name: Install libsndfile on Ubuntu
+      shell: bash
+      run: sudo apt-get install -y libsndfile-dev
+      if: startsWith(runner.os, 'Linux') == true
+
     - name: Setup Conda
       uses: s-weigand/setup-conda@v1
 
diff --git a/.github/workflows/develop_install.yml b/.github/workflows/develop_install.yml
index 39801e0dd..be7a4a110 100644
--- a/.github/workflows/develop_install.yml
+++ b/.github/workflows/develop_install.yml
@@ -17,21 +17,28 @@ jobs:
         pytorch_version: ['1.1.0', '1.2.0', '1.3.0', '1.3.1', '1.4.0']
         platform: ['windows-latest', 'ubuntu-latest', 'macos-latest']
 
+
         exclude:
           - platform: 'windows-latest'
             conda_python_version: '3.6'
+
           - platform: 'macos-latest'
             pytorch_version: '1.1.0'
+
           - pytorch_version: '1.1.0'
             conda_python_version: '3.8'
+
           - pytorch_version: '1.2.0'
             conda_python_version: '3.8'
+
           - pytorch_version: '1.3.0'
             conda_python_version: '3.8'
+
           - pytorch_version: '1.3.1'
             conda_python_version: '3.8'
 
 
+
     steps:
 
     - name: Checkout repo
@@ -72,6 +79,11 @@ jobs:
         mkdir $CONDA_PKGS_DIRS
         mkdir $PYTORCH_PKG_CACHE_PATH
 
+    - name: Install libsndfile on Ubuntu
+      shell: bash
+      run: sudo apt-get install -y libsndfile-dev
+      if: startsWith(runner.os, 'Linux') == true
+
     - name: Setup Conda
       uses: s-weigand/setup-conda@v1
 
@@ -115,9 +127,14 @@ jobs:
         rm -R ${{ steps.cache-dirs.outputs.pytorch_pkg_cache_path }}/pytorch*
       if: steps.pytorch-cache.outputs.cache-hit == 'true'
 
-    - name: Run Nox session for testing develop install and imports
+    - name: Run Nox session for testing brevitas develop install and imports
       shell: bash
-      run: nox --verbose --session tests_install_develop-${{ matrix.conda_python_version }}\(\pytorch_${{ matrix.pytorch_version }}\)
+      run: nox -v -s tests_brevitas_install_dev-${{ matrix.conda_python_version }}\(\pytorch_${{ matrix.pytorch_version }}\)
+
+    - name: Run Nox session for testing brevitas_examples develop install and imports
+      shell: bash
+      run: nox -v -s tests_brevitas_examples_install_dev-${{ matrix.conda_python_version }}\(\pytorch_${{ matrix.pytorch_version }}\)
+
 
 
     - name: Remove tarballs before caching Pytorch deps pkgs
diff --git a/.github/workflows/gen_github_actions.py b/.github/workflows/gen_github_actions.py
index 58ff73fa9..40cb54b07 100644
--- a/.github/workflows/gen_github_actions.py
+++ b/.github/workflows/gen_github_actions.py
@@ -42,15 +42,23 @@
 
 PYTEST_MATRIX_EXTRA = od([('jit_status', list(JIT_STATUSES))])
 
-PYTEST_STEP_LIST = [od([
-    ('name', 'Run Nox session for pytest'),
-    ('shell', 'bash'),
-    ('run', 'nox --verbose --session tests_cpu-${{ matrix.conda_python_version }}\(${{ matrix.jit_status }}\,\ pytorch_${{ matrix.pytorch_version }}\)')])]
-
-TEST_INSTALL_DEVELOP_STEP_LIST = [od([
-    ('name', 'Run Nox session for testing develop install and imports'),
-    ('shell', 'bash'),
-    ('run', 'nox --verbose --session tests_install_develop-${{ matrix.conda_python_version }}\(\pytorch_${{ matrix.pytorch_version }}\)')])]
+PYTEST_STEP_LIST = [
+    od([
+        ('name', 'Run Nox session for pytest'),
+        ('shell', 'bash'),
+        ('run', 'nox -v -s tests_brevitas_cpu-${{ matrix.conda_python_version }}\(${{ matrix.jit_status }}\,\ pytorch_${{ matrix.pytorch_version }}\)')
+    ])]
+
+TEST_INSTALL_DEV_STEP_LIST = [
+    od([
+        ('name', 'Run Nox session for testing brevitas develop install and imports'),
+        ('shell', 'bash'),
+        ('run', 'nox -v -s tests_brevitas_install_dev-${{ matrix.conda_python_version }}\(\pytorch_${{ matrix.pytorch_version }}\)')]),
+    od([
+        ('name', 'Run Nox session for testing brevitas_examples develop install and imports'),
+        ('shell', 'bash'),
+        ('run', 'nox -v -s tests_brevitas_examples_install_dev-${{ matrix.conda_python_version }}\(\pytorch_${{ matrix.pytorch_version }}\)')
+    ])]
 
 
 # whitespaces to indent generated portions of output yaml
@@ -90,6 +98,7 @@ def dict_str(d, quote_val, indent_first):
                 repr += f"{name}: {val}\n"
         if indent_first:
             repr = indent(repr, RELATIVE_INDENT*' ', predicate=lambda line: not first_line_prefix in line)
+        repr += '\n'
         return repr
 
     def gen_yaml(self, output_path):
@@ -122,7 +131,7 @@ def gen_test_develop_install_yml():
         'Test develop install',
         EXCLUDE_LIST,
         MATRIX,
-        TEST_INSTALL_DEVELOP_STEP_LIST)
+        TEST_INSTALL_DEV_STEP_LIST)
     test_develop_install.gen_yaml(DEVELOP_INSTALL_YML)
 
 
diff --git a/.github/workflows/pytest.yml b/.github/workflows/pytest.yml
index ffc5875cd..6cdfdf8f5 100644
--- a/.github/workflows/pytest.yml
+++ b/.github/workflows/pytest.yml
@@ -18,23 +18,31 @@ jobs:
         platform: ['windows-latest', 'ubuntu-latest', 'macos-latest']
         jit_status: ['jit_enabled', 'jit_disabled']
 
+
         exclude:
           - platform: 'windows-latest'
             conda_python_version: '3.6'
+
           - platform: 'macos-latest'
             pytorch_version: '1.1.0'
+
           - pytorch_version: '1.1.0'
             conda_python_version: '3.8'
+
           - pytorch_version: '1.2.0'
             conda_python_version: '3.8'
+
           - pytorch_version: '1.3.0'
             conda_python_version: '3.8'
+
           - pytorch_version: '1.3.1'
             conda_python_version: '3.8'
+
           - pytorch_version: '1.1.0'
             jit_status: 'jit_disabled'
 
 
+
     steps:
 
     - name: Checkout repo
@@ -75,6 +83,11 @@ jobs:
         mkdir $CONDA_PKGS_DIRS
         mkdir $PYTORCH_PKG_CACHE_PATH
 
+    - name: Install libsndfile on Ubuntu
+      shell: bash
+      run: sudo apt-get install -y libsndfile-dev
+      if: startsWith(runner.os, 'Linux') == true
+
     - name: Setup Conda
       uses: s-weigand/setup-conda@v1
 
@@ -120,7 +133,8 @@ jobs:
 
     - name: Run Nox session for pytest
       shell: bash
-      run: nox --verbose --session tests_cpu-${{ matrix.conda_python_version }}\(${{ matrix.jit_status }}\,\ pytorch_${{ matrix.pytorch_version }}\)
+      run: nox -v -s tests_brevitas_cpu-${{ matrix.conda_python_version }}\(${{ matrix.jit_status }}\,\ pytorch_${{ matrix.pytorch_version }}\)
+
 
 
     - name: Remove tarballs before caching Pytorch deps pkgs
diff --git a/brevitas_examples/__init__.py b/brevitas_examples/__init__.py
new file mode 100644
index 000000000..e69de29bb
diff --git a/examples/imagenet_classification/README.md b/brevitas_examples/imagenet_classification/README.md
similarity index 94%
rename from examples/imagenet_classification/README.md
rename to brevitas_examples/imagenet_classification/README.md
index 46d93de06..a63367265 100644
--- a/examples/imagenet_classification/README.md
+++ b/brevitas_examples/imagenet_classification/README.md
@@ -16,12 +16,12 @@ Below in the table is a list of example pretrained models made available for ref
 To evaluate a pretrained quantized model on ImageNet:
  
  - Make sure you have Brevitas installed and the ImageNet dataset in a Pytorch friendly format (following this [script](https://raw.githubusercontent.com/soumith/imagenetloader.torch/master/valprep.sh)).
- - Pass the corresponding cfg .ini file as an input to the evaluation script. The required checkpoint will be downloaded automatically. 
+ - Pass the name of the model as an input to the evaluation script. The required checkpoint will be downloaded automatically. 
  
  For example, for *quant_mobilenet_v1_4b* evaluated on GPU 0:
 
 ```
-python imagenet_val.py --imagenet-dir /path/to/imagenet --model-cfg ./cfg/quant_mobilenet_v1_4b.ini --gpu 0
+brevitas_imagenet_val --imagenet-dir /path/to/imagenet --model quant_mobilenet_v1_4b --gpu 0 --pretrained
 ```
 
 ## MobileNet V1
diff --git a/brevitas_examples/imagenet_classification/__init__.py b/brevitas_examples/imagenet_classification/__init__.py
new file mode 100644
index 000000000..cf4f59d6c
--- /dev/null
+++ b/brevitas_examples/imagenet_classification/__init__.py
@@ -0,0 +1 @@
+from .models import *
\ No newline at end of file
diff --git a/brevitas_examples/imagenet_classification/cfg/__init__.py b/brevitas_examples/imagenet_classification/cfg/__init__.py
new file mode 100644
index 000000000..e69de29bb
diff --git a/examples/imagenet_classification/cfg/quant_mobilenet_v1_4b.ini b/brevitas_examples/imagenet_classification/cfg/quant_mobilenet_v1_4b.ini
similarity index 100%
rename from examples/imagenet_classification/cfg/quant_mobilenet_v1_4b.ini
rename to brevitas_examples/imagenet_classification/cfg/quant_mobilenet_v1_4b.ini
diff --git a/examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_4b.ini b/brevitas_examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_4b.ini
similarity index 100%
rename from examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_4b.ini
rename to brevitas_examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_4b.ini
diff --git a/examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_4b5b.ini b/brevitas_examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_4b5b.ini
similarity index 100%
rename from examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_4b5b.ini
rename to brevitas_examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_4b5b.ini
diff --git a/examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_hadamard_4b.ini b/brevitas_examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_hadamard_4b.ini
similarity index 100%
rename from examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_hadamard_4b.ini
rename to brevitas_examples/imagenet_classification/cfg/quant_proxylessnas_mobile14_hadamard_4b.ini
diff --git a/examples/imagenet_classification/imagenet_val.py b/brevitas_examples/imagenet_classification/imagenet_val.py
similarity index 81%
rename from examples/imagenet_classification/imagenet_val.py
rename to brevitas_examples/imagenet_classification/imagenet_val.py
index a256aacd0..17cde7772 100644
--- a/examples/imagenet_classification/imagenet_val.py
+++ b/brevitas_examples/imagenet_classification/imagenet_val.py
@@ -12,17 +12,15 @@
 import torchvision.transforms as transforms
 import torchvision.datasets as datasets
 
-from models import *
+from .models import model_with_cfg
 
 SEED = 123456
 
-models = {'quant_mobilenet_v1': quant_mobilenet_v1,
-          'quant_proxylessnas_mobile14': quant_proxylessnas_mobile14}
-
 
 parser = argparse.ArgumentParser(description='PyTorch ImageNet Validation')
 parser.add_argument('--imagenet-dir', help='path to folder containing Imagenet val folder')
-parser.add_argument('--model-cfg', type=str, help='Path to pretrained model .ini configuration file')
+parser.add_argument('--model', type=str, default='quant_mobilenet_v1_4b', help='Name of the model')
+parser.add_argument('--pretrained', action='store_true', help='Load pretrained checkpoint')
 parser.add_argument('--workers', default=4, type=int, help='number of data loading workers')
 parser.add_argument('--batch-size', default=256, type=int, help='Minibatch size')
 parser.add_argument('--gpu', default=None, type=int, help='GPU id to use.')
@@ -33,28 +31,13 @@ def main():
     random.seed(SEED)
     torch.manual_seed(SEED)
 
-    assert os.path.exists(args.model_cfg)
-    cfg = configparser.ConfigParser()
-    cfg.read(args.model_cfg)
-    arch = cfg.get('MODEL', 'ARCH')
-
-    model = models[arch](cfg)
+    model, cfg = model_with_cfg(args.model, args.pretrained)
 
     if args.gpu is not None:
         torch.cuda.set_device(args.gpu)
         model = model.cuda(args.gpu)
         cudnn.benchmark = True
 
-    pretrained_url = cfg.get('MODEL', 'PRETRAINED_URL')
-    print("=> Loading checkpoint from:'{}'".format(pretrained_url))
-    if args.gpu is None:
-        checkpoint = torch.hub.load_state_dict_from_url(pretrained_url)
-    else:
-        # Map model to be loaded to specified single gpu.
-        loc = 'cuda:{}'.format(args.gpu)
-        checkpoint = torch.hub.load_state_dict_from_url(pretrained_url, map_location=loc)
-    model.load_state_dict(checkpoint, strict=True)
-
     valdir = os.path.join(args.imagenet_dir, 'val')
     mean = [float(cfg.get('PREPROCESS', 'MEAN_0')), float(cfg.get('PREPROCESS', 'MEAN_1')),
             float(cfg.get('PREPROCESS', 'MEAN_2'))]
diff --git a/brevitas_examples/imagenet_classification/models/__init__.py b/brevitas_examples/imagenet_classification/models/__init__.py
new file mode 100644
index 000000000..2319f2aff
--- /dev/null
+++ b/brevitas_examples/imagenet_classification/models/__init__.py
@@ -0,0 +1,48 @@
+import os
+from configparser import ConfigParser
+
+from torch import hub
+
+from .mobilenetv1 import *
+from .vgg import *
+from .proxylessnas import *
+
+model_impl = {
+    'quant_mobilenet_v1': quant_mobilenet_v1,
+    'quant_proxylessnas_mobile14': quant_proxylessnas_mobile14
+}
+
+
+def model_with_cfg(name, pretrained):
+    cfg = ConfigParser()
+    current_dir = os.path.dirname(os.path.abspath(__file__))
+    config_path = os.path.join(current_dir, '..', 'cfg', name + '.ini')
+    assert os.path.exists(config_path)
+    cfg.read(config_path)
+    arch = cfg.get('MODEL', 'ARCH')
+    model = model_impl[arch](cfg)
+    if pretrained:
+        checkpoint = cfg.get('MODEL', 'PRETRAINED_URL')
+        state_dict = hub.load_state_dict_from_url(checkpoint, progress=True, map_location='cpu')
+        model.load_state_dict(state_dict, strict=True)
+    return model, cfg
+
+
+def quant_mobilenet_v1_4b(pretrained=True):
+    model, _ = model_with_cfg('quant_mobilenet_v1_4b', pretrained)
+    return model
+
+
+def quant_proxylessnas_mobile14_4b(pretrained=True):
+    model, _ = model_with_cfg('quant_proxylessnas_mobile14_4b', pretrained)
+    return model
+
+
+def quant_proxylessnas_mobile14_4b5b(pretrained=True):
+    model, _ = model_with_cfg('quant_proxylessnas_mobile14_4b5b', pretrained)
+    return model
+
+
+def quant_proxylessnas_mobile14_hadamard_4b(pretrained=True):
+    model, _ = model_with_cfg('quant_proxylessnas_mobile14_hadamard_4b', pretrained)
+    return model
\ No newline at end of file
diff --git a/examples/imagenet_classification/models/common.py b/brevitas_examples/imagenet_classification/models/common.py
similarity index 99%
rename from examples/imagenet_classification/models/common.py
rename to brevitas_examples/imagenet_classification/models/common.py
index d2e56bb17..2acb53ca1 100644
--- a/examples/imagenet_classification/models/common.py
+++ b/brevitas_examples/imagenet_classification/models/common.py
@@ -4,6 +4,7 @@
 from brevitas.core.scaling import ScalingImplType
 from brevitas.core.stats import StatsOp
 
+
 QUANT_TYPE = QuantType.INT
 SCALING_MIN_VAL = 2e-16
 
@@ -151,3 +152,6 @@ def make_hadamard_classifier(in_channels,
     return qnn.HadamardClassifier(in_channels=in_channels,
                                   out_channels=out_channels,
                                   fixed_scale=fixed_scale)
+
+
+
diff --git a/examples/imagenet_classification/models/mobilenetv1.py b/brevitas_examples/imagenet_classification/models/mobilenetv1.py
similarity index 98%
rename from examples/imagenet_classification/models/mobilenetv1.py
rename to brevitas_examples/imagenet_classification/models/mobilenetv1.py
index 2b4c7a2b0..3f189ee8a 100644
--- a/examples/imagenet_classification/models/mobilenetv1.py
+++ b/brevitas_examples/imagenet_classification/models/mobilenetv1.py
@@ -33,7 +33,7 @@
 
 from brevitas.quant_tensor import pack_quant_tensor
 
-from .common import make_quant_conv2d, make_quant_linear, make_quant_relu, make_quant_avg_pool
+from .common import *
 
 
 FIRST_LAYER_BIT_WIDTH = 8
@@ -170,3 +170,6 @@ def quant_mobilenet_v1(cfg):
                     first_stage_stride=first_stage_stride,
                     bit_width=bit_width)
     return net
+
+
+
diff --git a/examples/imagenet_classification/models/proxylessnas.py b/brevitas_examples/imagenet_classification/models/proxylessnas.py
similarity index 100%
rename from examples/imagenet_classification/models/proxylessnas.py
rename to brevitas_examples/imagenet_classification/models/proxylessnas.py
diff --git a/examples/imagenet_classification/models/vgg.py b/brevitas_examples/imagenet_classification/models/vgg.py
similarity index 100%
rename from examples/imagenet_classification/models/vgg.py
rename to brevitas_examples/imagenet_classification/models/vgg.py
diff --git a/examples/speech_to_text/README.md b/brevitas_examples/speech_to_text/README.md
similarity index 80%
rename from examples/speech_to_text/README.md
rename to brevitas_examples/speech_to_text/README.md
index 4308553d8..c02c1de09 100644
--- a/examples/speech_to_text/README.md
+++ b/brevitas_examples/speech_to_text/README.md
@@ -15,19 +15,20 @@ It is highly recommended to setup a virtual environment.
 
 Download and pre-process the LibriSpeech dataset with the following command:
 ```
-python utilities/get_librispeech_data.py --data_root=/path/to/validation/folder --data_set=DEV_OTHER
+brevitas_quartznet_preprocess --data_root=/path/to/validation/folder --data_set=DEV_OTHER
 ```
 
 To evaluate a pretrained quantized model on LibriSpeech:
 
  - Install pytorch from the [Pytorch Website](https://pytorch.org/), and Cython with the following command:
  `python install --upgrade cython`
- - Install  the Quartznet requirements with `pip install requirements.txt`
- - Make sure you have Brevitas installed
- - Pass the corresponding cfg .ini file as an input to the evaluation script. The required checkpoint will be downloaded automatically. 
+ - Install SoX (this [guide](https://at.projects.genivi.org/wiki/display/PROJ/Installation+of+SoX+on+different+Platforms)
+ may be helpful)
+ - After cloning the repository, install Brevitas and QuartzNet requirements with `pip install .[stt]`
+ - Pass the name of the model as an input to the evaluation script. The required checkpoint will be downloaded automatically. 
  
  For example, for the evaluation on GPU 0:
 
 ```
-python quartznet_val.py --input-folder /path/to/validation/folder --model-cfg cfg/quant_quartznet_pertensorscaling_8b.ini --gpu 0
+brevitas_quartznet_val --input-folder /path/to/validation/folder --model quant_quartznet_pertensorscaling_8b --gpu 0 --pretrained
 ```
diff --git a/brevitas_examples/speech_to_text/__init__.py b/brevitas_examples/speech_to_text/__init__.py
new file mode 100644
index 000000000..8d6fb3970
--- /dev/null
+++ b/brevitas_examples/speech_to_text/__init__.py
@@ -0,0 +1 @@
+from .quartznet import *
\ No newline at end of file
diff --git a/brevitas_examples/speech_to_text/cfg/__init__.py b/brevitas_examples/speech_to_text/cfg/__init__.py
new file mode 100644
index 000000000..e69de29bb
diff --git a/examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_4b.ini b/brevitas_examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_4b.ini
similarity index 92%
rename from examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_4b.ini
rename to brevitas_examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_4b.ini
index 5f2f82e03..62ea54a6c 100644
--- a/examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_4b.ini
+++ b/brevitas_examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_4b.ini
@@ -1,6 +1,6 @@
 [MODEL]
 ARCH: quartznet
-TOPOLOGY_FILE: cfg/topology/quartznet15x5.yaml
+TOPOLOGY_FILE: quartznet15x5.yaml
 PRETRAINED_ENCODER_URL: https://github.com/Xilinx/brevitas/releases/download/quant_quartznet_4b-r0/quant_quartznet_encoder_4b-0a46a232.pth
 PRETRAINED_DECODER_URL: https://github.com/Xilinx/brevitas/releases/download/quant_quartznet_4b-r0/quant_quartznet_decoder_4b-bcbf8c7b.pth
 
diff --git a/examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_8b.ini b/brevitas_examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_8b.ini
similarity index 92%
rename from examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_8b.ini
rename to brevitas_examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_8b.ini
index 58005ebbb..2aa28a511 100644
--- a/examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_8b.ini
+++ b/brevitas_examples/speech_to_text/cfg/quant_quartznet_perchannelscaling_8b.ini
@@ -1,6 +1,6 @@
 [MODEL]
 ARCH: quartznet
-TOPOLOGY_FILE: cfg/topology/quartznet15x5.yaml
+TOPOLOGY_FILE: quartznet15x5.yaml
 PRETRAINED_ENCODER_URL: https://github.com/Xilinx/brevitas/releases/download/quant_quartznet_8b-r0/quant_quartznet_encoder_8b-50f12b4b.pth
 PRETRAINED_DECODER_URL: https://github.com/Xilinx/brevitas/releases/download/quant_quartznet_8b-r0/quant_quartznet_decoder_8b-af09651c.pth
 
diff --git a/examples/speech_to_text/cfg/quant_quartznet_pertensorscaling_8b.ini b/brevitas_examples/speech_to_text/cfg/quant_quartznet_pertensorscaling_8b.ini
similarity index 92%
rename from examples/speech_to_text/cfg/quant_quartznet_pertensorscaling_8b.ini
rename to brevitas_examples/speech_to_text/cfg/quant_quartznet_pertensorscaling_8b.ini
index 8d025e9c7..c7f50261b 100644
--- a/examples/speech_to_text/cfg/quant_quartznet_pertensorscaling_8b.ini
+++ b/brevitas_examples/speech_to_text/cfg/quant_quartznet_pertensorscaling_8b.ini
@@ -1,6 +1,6 @@
 [MODEL]
 ARCH: quartznet
-TOPOLOGY_FILE: cfg/topology/quartznet15x5.yaml
+TOPOLOGY_FILE: quartznet15x5.yaml
 PRETRAINED_ENCODER_URL: https://github.com/Xilinx/brevitas/releases/download/quant_quartznet_8b-r0/quant_quartznet_encoder_8b-50f12b4b.pth
 PRETRAINED_DECODER_URL: https://github.com/Xilinx/brevitas/releases/download/quant_quartznet_8b-r0/quant_quartznet_decoder_8b-af09651c.pth
 
diff --git a/brevitas_examples/speech_to_text/cfg/topology/__init__.py b/brevitas_examples/speech_to_text/cfg/topology/__init__.py
new file mode 100644
index 000000000..e69de29bb
diff --git a/examples/speech_to_text/cfg/topology/quartznet15x5.yaml b/brevitas_examples/speech_to_text/cfg/topology/quartznet15x5.yaml
similarity index 100%
rename from examples/speech_to_text/cfg/topology/quartznet15x5.yaml
rename to brevitas_examples/speech_to_text/cfg/topology/quartznet15x5.yaml
diff --git a/brevitas_examples/speech_to_text/get_librispeech_data.py b/brevitas_examples/speech_to_text/get_librispeech_data.py
new file mode 100644
index 000000000..495bb1665
--- /dev/null
+++ b/brevitas_examples/speech_to_text/get_librispeech_data.py
@@ -0,0 +1,148 @@
+# Adapted from https://github.com/NVIDIA/NeMo/tree/r0.9
+# Copyright (C) 2020 Xilinx (Giuseppe Franco)
+# Copyright (C) 2019 NVIDIA CORPORATION.
+#
+# 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.
+
+import argparse
+import fnmatch
+import json
+import os
+import subprocess
+import tarfile
+import urllib.request
+
+from sox import Transformer
+from tqdm import tqdm
+
+parser = argparse.ArgumentParser(description='LibriSpeech Data download')
+parser.add_argument("--data_root", required=True, default=None, type=str)
+parser.add_argument("--data_sets", default="dev_clean", type=str)
+args = parser.parse_args()
+
+URLS = {
+    'TRAIN_CLEAN_100': ("http://www.openslr.org/resources/12/train-clean-100.tar.gz"),
+    'TRAIN_CLEAN_360': ("http://www.openslr.org/resources/12/train-clean-360.tar.gz"),
+    'TRAIN_OTHER_500': ("http://www.openslr.org/resources/12/train-other-500.tar.gz"),
+    'DEV_CLEAN': "http://www.openslr.org/resources/12/dev-clean.tar.gz",
+    'DEV_OTHER': "http://www.openslr.org/resources/12/dev-other.tar.gz",
+    'TEST_CLEAN': "http://www.openslr.org/resources/12/test-clean.tar.gz",
+    'TEST_OTHER': "http://www.openslr.org/resources/12/test-other.tar.gz",
+}
+
+
+def __maybe_download_file(destination: str, source: str):
+    """
+    Downloads source to destination if it doesn't exist.
+    If exists, skips download
+    Args:
+        destination: local filepath
+        source: url of resource
+
+    Returns:
+
+    """
+    source = URLS[source]
+    if not os.path.exists(destination):
+        print("{0} does not exist. Downloading ...".format(destination))
+        urllib.request.urlretrieve(source, filename=destination + '.tmp')
+        os.rename(destination + '.tmp', destination)
+        print("Downloaded {0}.".format(destination))
+    else:
+        print("Destination {0} exists. Skipping.".format(destination))
+    return destination
+
+
+def __extract_file(filepath: str, data_dir: str):
+    try:
+        tar = tarfile.open(filepath)
+        tar.extractall(data_dir)
+        tar.close()
+    except Exception:
+        print('Not extracting. Maybe already there?')
+
+
+def __process_data(data_folder: str, dst_folder: str, manifest_file: str):
+    """
+    Converts flac to wav and build manifests's json
+    Args:
+        data_folder: source with flac files
+        dst_folder: where wav files will be stored
+        manifest_file: where to store manifest
+
+    Returns:
+
+    """
+
+    if not os.path.exists(dst_folder):
+        os.makedirs(dst_folder)
+
+    files = []
+    entries = []
+
+    for root, dirnames, filenames in os.walk(data_folder):
+        for filename in fnmatch.filter(filenames, '*.trans.txt'):
+            files.append((os.path.join(root, filename), root))
+
+    for transcripts_file, root in tqdm(files):
+        with open(transcripts_file, encoding="utf-8") as fin:
+            for line in fin:
+                id, text = line[: line.index(" ")], line[line.index(" ") + 1 :]
+                transcript_text = text.lower().strip()
+
+                # Convert FLAC file to WAV
+                flac_file = os.path.join(root, id + ".flac")
+                wav_file = os.path.join(dst_folder, id + ".wav")
+                if not os.path.exists(wav_file):
+                    Transformer().build(flac_file, wav_file)
+                # check duration
+                duration = subprocess.check_output("soxi -D {0}".format(wav_file), shell=True)
+
+                entry = dict()
+                entry['audio_filepath'] = os.path.abspath(wav_file)
+                entry['duration'] = float(duration)
+                entry['text'] = transcript_text
+                entries.append(entry)
+
+    with open(manifest_file, 'w') as fout:
+        for m in entries:
+            fout.write(json.dumps(m) + '\n')
+
+
+def main():
+    data_root = args.data_root
+    data_sets = args.data_sets
+
+    if data_sets == "ALL":
+        data_sets = "dev_clean,dev_other,train_clean_100,train_clean_360,train_other_500,test_clean,test_other"
+
+    for data_set in data_sets.split(','):
+        print("\n\nWorking on: {0}".format(data_set))
+        filepath = os.path.join(data_root, data_set + ".tar.gz")
+        print("Getting {0}".format(data_set))
+        __maybe_download_file(filepath, data_set.upper())
+        print("Extracting {0}".format(data_set))
+        __extract_file(filepath, data_root)
+        print("Processing {0}".format(data_set))
+        __process_data(
+            os.path.join(os.path.join(data_root, "LibriSpeech"), data_set.replace("_", "-"),),
+            os.path.join(os.path.join(data_root, "LibriSpeech"), data_set.replace("_", "-"),) + "-processed",
+            os.path.join(data_root, data_set + ".json"),
+        )
+    print('Done!')
+
+
+if __name__ == "__main__":
+    main()
\ No newline at end of file
diff --git a/brevitas_examples/speech_to_text/quartznet/__init__.py b/brevitas_examples/speech_to_text/quartznet/__init__.py
new file mode 100644
index 000000000..6c894e610
--- /dev/null
+++ b/brevitas_examples/speech_to_text/quartznet/__init__.py
@@ -0,0 +1,80 @@
+# Adapted from https://github.com/NVIDIA/NeMo/blob/r0.9/collections/nemo_asr/
+# Copyright (C) 2020 Xilinx (Giuseppe Franco)
+# Copyright (C) 2019 NVIDIA CORPORATION.
+#
+# 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.
+
+
+from .data_layer import (
+        AudioToTextDataLayer)
+from .greedy_ctc_decoder import GreedyCTCDecoder
+from .quartznet import quartznet
+from .losses import CTCLossNM
+from .helpers import *
+
+import os
+from configparser import ConfigParser
+from ruamel.yaml import YAML
+from torch import hub
+
+__all__ = ['AudioToTextDataLayer',
+           'quartznet',
+           'quant_quartznet_perchannelscaling_4b',
+           'quant_quartznet_perchannelscaling_8b',
+           'quant_quartznet_pertensorscaling_8b']
+
+name = "quarznet_release"
+model_impl = {
+    'quartznet': quartznet,
+}
+
+
+def model_with_cfg(name, pretrained):
+    cfg = ConfigParser()
+    current_dir = os.path.dirname(os.path.abspath(__file__))
+    config_path = os.path.join(current_dir, '..', 'cfg', name + '.ini')
+    assert os.path.exists(config_path)
+    cfg.read(config_path)
+    arch = cfg.get('MODEL', 'ARCH')
+    topology_file = cfg.get('MODEL', 'TOPOLOGY_FILE')
+    topology_path = os.path.join(current_dir, '..', 'cfg', 'topology', topology_file)
+    yaml = YAML(typ="safe")
+    with open(topology_path) as f:
+        quartnzet_params = yaml.load(f)
+    model = model_impl[arch](cfg, quartnzet_params)
+    if pretrained:
+        pretrained_encoder_url = cfg.get('MODEL', 'PRETRAINED_ENCODER_URL')
+        pretrained_decoder_url = cfg.get('MODEL', 'PRETRAINED_DECODER_URL')
+        print("=> Loading encoder checkpoint from:'{}'".format(pretrained_encoder_url))
+        print("=> Loading decoder checkpoint from:'{}'".format(pretrained_decoder_url))
+        checkpoint_enc = torch.hub.load_state_dict_from_url(pretrained_encoder_url, progress=True, map_location='cpu')
+        checkpoint_dec = torch.hub.load_state_dict_from_url(pretrained_decoder_url, progress=True, map_location='cpu')
+        model.restore_checkpoints(checkpoint_enc, checkpoint_dec)
+    return model, cfg
+
+
+def quant_quartznet_perchannelscaling_4b(pretrained=True):
+    model, _ = model_with_cfg('quant_quartznet_perchannelscaling_4b', pretrained)
+    return model
+
+
+def quant_quartznet_perchannelscaling_8b(pretrained=True):
+    model, _ = model_with_cfg('quant_quartznet_perchannelscaling_8b', pretrained)
+    return model
+
+
+def quant_quartznet_pertensorscaling_8b(pretrained=True):
+    model, _ = model_with_cfg('quant_quartznet_pertensorscaling_8b', pretrained)
+    return model
diff --git a/examples/speech_to_text/quartznet/audio_preprocessing.py b/brevitas_examples/speech_to_text/quartznet/audio_preprocessing.py
similarity index 100%
rename from examples/speech_to_text/quartznet/audio_preprocessing.py
rename to brevitas_examples/speech_to_text/quartznet/audio_preprocessing.py
diff --git a/examples/speech_to_text/quartznet/data_layer.py b/brevitas_examples/speech_to_text/quartznet/data_layer.py
similarity index 95%
rename from examples/speech_to_text/quartznet/data_layer.py
rename to brevitas_examples/speech_to_text/quartznet/data_layer.py
index 4594cb1d5..21bb852b1 100644
--- a/examples/speech_to_text/quartznet/data_layer.py
+++ b/brevitas_examples/speech_to_text/quartznet/data_layer.py
@@ -25,10 +25,7 @@
 import torch
 import torch.nn as nn
 
-# from nemo.backends.pytorch import DataLayerNM
-# from nemo.core import DeviceType
-# from nemo.core.neural_types import *
-from .parts.dataset import (AudioDataset, seq_collate_fn)  # , KaldiFeatureDataset, TranscriptDataset)
+from .parts.dataset import (AudioDataset, seq_collate_fn)
 from .parts.features import WaveformFeaturizer
 
 def pad_to(x, k=8):
@@ -117,7 +114,7 @@ def __init__(
             load_audio=True,
             drop_last=False,
             shuffle=True,
-            num_workers=0,
+            num_workers=4,
             placement='cpu',
             # perturb_config=None,
             **kwargs
diff --git a/examples/speech_to_text/quartznet/greedy_ctc_decoder.py b/brevitas_examples/speech_to_text/quartznet/greedy_ctc_decoder.py
similarity index 100%
rename from examples/speech_to_text/quartznet/greedy_ctc_decoder.py
rename to brevitas_examples/speech_to_text/quartznet/greedy_ctc_decoder.py
diff --git a/examples/speech_to_text/quartznet/helpers.py b/brevitas_examples/speech_to_text/quartznet/helpers.py
similarity index 100%
rename from examples/speech_to_text/quartznet/helpers.py
rename to brevitas_examples/speech_to_text/quartznet/helpers.py
diff --git a/examples/speech_to_text/quartznet/losses.py b/brevitas_examples/speech_to_text/quartznet/losses.py
similarity index 100%
rename from examples/speech_to_text/quartznet/losses.py
rename to brevitas_examples/speech_to_text/quartznet/losses.py
diff --git a/examples/speech_to_text/quartznet/metrics.py b/brevitas_examples/speech_to_text/quartznet/metrics.py
similarity index 100%
rename from examples/speech_to_text/quartznet/metrics.py
rename to brevitas_examples/speech_to_text/quartznet/metrics.py
diff --git a/examples/speech_to_text/quartznet/parts/__init__.py b/brevitas_examples/speech_to_text/quartznet/parts/__init__.py
similarity index 100%
rename from examples/speech_to_text/quartznet/parts/__init__.py
rename to brevitas_examples/speech_to_text/quartznet/parts/__init__.py
diff --git a/examples/speech_to_text/quartznet/parts/cleaners.py b/brevitas_examples/speech_to_text/quartznet/parts/cleaners.py
similarity index 100%
rename from examples/speech_to_text/quartznet/parts/cleaners.py
rename to brevitas_examples/speech_to_text/quartznet/parts/cleaners.py
diff --git a/examples/speech_to_text/quartznet/parts/common.py b/brevitas_examples/speech_to_text/quartznet/parts/common.py
similarity index 100%
rename from examples/speech_to_text/quartznet/parts/common.py
rename to brevitas_examples/speech_to_text/quartznet/parts/common.py
diff --git a/examples/speech_to_text/quartznet/parts/dataset.py b/brevitas_examples/speech_to_text/quartznet/parts/dataset.py
similarity index 99%
rename from examples/speech_to_text/quartznet/parts/dataset.py
rename to brevitas_examples/speech_to_text/quartznet/parts/dataset.py
index 283770fb8..a6b60a119 100644
--- a/examples/speech_to_text/quartznet/parts/dataset.py
+++ b/brevitas_examples/speech_to_text/quartznet/parts/dataset.py
@@ -23,9 +23,6 @@
 # OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 # SOFTWARE.
 
-import os
-import pandas as pd
-import string
 import torch
 from torch.utils.data import Dataset
 
diff --git a/examples/speech_to_text/quartznet/parts/features.py b/brevitas_examples/speech_to_text/quartznet/parts/features.py
similarity index 100%
rename from examples/speech_to_text/quartznet/parts/features.py
rename to brevitas_examples/speech_to_text/quartznet/parts/features.py
diff --git a/examples/speech_to_text/quartznet/parts/manifest.py b/brevitas_examples/speech_to_text/quartznet/parts/manifest.py
similarity index 100%
rename from examples/speech_to_text/quartznet/parts/manifest.py
rename to brevitas_examples/speech_to_text/quartznet/parts/manifest.py
diff --git a/examples/speech_to_text/quartznet/parts/perturb.py b/brevitas_examples/speech_to_text/quartznet/parts/perturb.py
similarity index 100%
rename from examples/speech_to_text/quartznet/parts/perturb.py
rename to brevitas_examples/speech_to_text/quartznet/parts/perturb.py
diff --git a/examples/speech_to_text/quartznet/parts/quartznet.py b/brevitas_examples/speech_to_text/quartznet/parts/quartznet.py
similarity index 100%
rename from examples/speech_to_text/quartznet/parts/quartznet.py
rename to brevitas_examples/speech_to_text/quartznet/parts/quartznet.py
diff --git a/examples/speech_to_text/quartznet/parts/segment.py b/brevitas_examples/speech_to_text/quartznet/parts/segment.py
similarity index 100%
rename from examples/speech_to_text/quartznet/parts/segment.py
rename to brevitas_examples/speech_to_text/quartznet/parts/segment.py
diff --git a/examples/speech_to_text/quartznet/parts/spectr_augment.py b/brevitas_examples/speech_to_text/quartznet/parts/spectr_augment.py
similarity index 100%
rename from examples/speech_to_text/quartznet/parts/spectr_augment.py
rename to brevitas_examples/speech_to_text/quartznet/parts/spectr_augment.py
diff --git a/examples/speech_to_text/quartznet/quartznet.py b/brevitas_examples/speech_to_text/quartznet/quartznet.py
similarity index 100%
rename from examples/speech_to_text/quartznet/quartznet.py
rename to brevitas_examples/speech_to_text/quartznet/quartznet.py
diff --git a/examples/speech_to_text/quartznet_val.py b/brevitas_examples/speech_to_text/quartznet_val.py
similarity index 69%
rename from examples/speech_to_text/quartznet_val.py
rename to brevitas_examples/speech_to_text/quartznet_val.py
index 99a8edb69..77a142f20 100644
--- a/examples/speech_to_text/quartznet_val.py
+++ b/brevitas_examples/speech_to_text/quartznet_val.py
@@ -7,22 +7,24 @@
 import random
 import torch
 
-from quartznet import AudioToTextDataLayer, quartznet
-from quartznet.helpers import word_error_rate, post_process_predictions, \
+from .quartznet import AudioToTextDataLayer
+from .quartznet.helpers import word_error_rate, post_process_predictions, \
     post_process_transcripts
 import torch.backends.cudnn as cudnn
 import brevitas.config
+from .quartznet import model_with_cfg
+
 brevitas.config.IGNORE_MISSING_KEYS = False
 SEED = 123456
 BATCH_SIZE = 64
 
 
-models = {'quartznet': quartznet}
 
 parser = argparse.ArgumentParser(description='Quartznet')
-parser.add_argument("--model-cfg", type=str, required=True)
 parser.add_argument("--input-folder", type=str, required=False)
 parser.add_argument("--gpu", type=int)
+parser.add_argument('--pretrained', action='store_true', default=True, help='Load pretrained checkpoint')
+parser.add_argument('--model', type=str, default='quant_quartznet_perchannelscaling_4b', help='Name of the model')
 
 
 def main():
@@ -31,19 +33,15 @@ def main():
 
     args = parser.parse_args()
 
-    assert os.path.exists(args.model_cfg)
-    cfg = configparser.ConfigParser()
-    cfg.read(args.model_cfg)
+    model, cfg = model_with_cfg(args.model, args.pretrained)
 
     topology_file = cfg.get('MODEL', 'TOPOLOGY_FILE')
-
+    current_dir = os.path.dirname(os.path.abspath(__file__))
+    topology_path = os.path.join(current_dir, 'cfg', 'topology', topology_file)
     yaml = YAML(typ="safe")
-    with open(topology_file) as f:
+    with open(topology_path) as f:
         quartnzet_params = yaml.load(f)
 
-    arch = cfg.get('MODEL', 'ARCH')
-
-    model = models[arch](cfg, quartnzet_params)
     vocab = quartnzet_params['labels']
     sample_rate = quartnzet_params['sample_rate']
 
@@ -79,28 +77,14 @@ def main():
     print('================================')
 
 
-
     if args.gpu is not None:
         torch.cuda.set_device(args.gpu)
         cudnn.benchmark = True
-
-    pretrained_encoder_url = cfg.get('MODEL', 'PRETRAINED_ENCODER_URL')
-    pretrained_decoder_url = cfg.get('MODEL', 'PRETRAINED_DECODER_URL')
-    print("=> Loading encoder checkpoint from:'{}'".format(pretrained_encoder_url))
-    print("=> Loading decoder checkpoint from:'{}'".format(pretrained_decoder_url))
-    if args.gpu is None:
-        loc = 'cpu'
-        checkpoint_enc = torch.hub.load_state_dict_from_url(pretrained_encoder_url)
-        checkpoint_dec = torch.hub.load_state_dict_from_url(pretrained_decoder_url)
-    else:
-        # Map model to be loaded to specified single gpu.
         loc = 'cuda:{}'.format(args.gpu)
-        checkpoint_enc = torch.hub.load_state_dict_from_url(pretrained_encoder_url, map_location=loc)
-        checkpoint_dec = torch.hub.load_state_dict_from_url(pretrained_decoder_url, map_location=loc)
-
-    model.restore_checkpoints(checkpoint_enc, checkpoint_dec)
+    else:
+        loc = 'cpu'
     model.to(loc)
-
+    
     predictions = []
     transcripts = []
     transcripts_len = []
diff --git a/examples/text_to_speech/README.md b/brevitas_examples/text_to_speech/README.md
similarity index 78%
rename from examples/text_to_speech/README.md
rename to brevitas_examples/text_to_speech/README.md
index df07c5240..aacf086f0 100644
--- a/examples/text_to_speech/README.md
+++ b/brevitas_examples/text_to_speech/README.md
@@ -19,13 +19,12 @@ find /path/to/LJSpeech/LJSpeech-1.1/wavs -type f | head -10 | xargs cp -t /path/
 
 To evaluate a pretrained quantized model on LJSpeech1.1:
 
- - Install  the MelGAN requirements with `pip install requirements.txt`
- - Make sure you have Brevitas installed
- - Preprocess the dataset with `python preprocess_datasets --model-cfg cfg/quant_melgan_8b --data-path /path/to/validation/folder`
- - Pass the corresponding cfg .ini file as an input to the evaluation script. The required checkpoint will be downloaded automatically. 
+ - After cloning the repository, install Brevitas and MelGAN requirements with `pip install .[tts]`
+ - Preprocess the dataset with `brevitas_melgan_preprocess --name quant_melgan_8b --data-path /path/to/validation/folder`
+ - Pass the name of the model as an input to the evaluation script. The required checkpoint will be downloaded automatically. 
  
  For example, for the evaluation on GPU 0:
 
 ```
-python melgan_val.py --input-folder /path/to/validation/folder --model-cfg cfg/quant_melgan_8b --gpu 0
+brevitas_melgan_val --input-folder /path/to/validation/folder --model quant_melgan_8b --gpu 0 --pretrained
 ```
diff --git a/examples/text_to_speech/MelGAN/__init__.py b/brevitas_examples/text_to_speech/__init__.py
similarity index 100%
rename from examples/text_to_speech/MelGAN/__init__.py
rename to brevitas_examples/text_to_speech/__init__.py
diff --git a/brevitas_examples/text_to_speech/cfg/__init__.py b/brevitas_examples/text_to_speech/cfg/__init__.py
new file mode 100644
index 000000000..e69de29bb
diff --git a/examples/text_to_speech/cfg/quant_melgan_8b.ini b/brevitas_examples/text_to_speech/cfg/quant_melgan_8b.ini
similarity index 100%
rename from examples/text_to_speech/cfg/quant_melgan_8b.ini
rename to brevitas_examples/text_to_speech/cfg/quant_melgan_8b.ini
diff --git a/brevitas_examples/text_to_speech/melgan/__init__.py b/brevitas_examples/text_to_speech/melgan/__init__.py
new file mode 100644
index 000000000..6bb4c5fff
--- /dev/null
+++ b/brevitas_examples/text_to_speech/melgan/__init__.py
@@ -0,0 +1,30 @@
+import os
+from configparser import ConfigParser
+
+from torch import hub
+
+from .melgan import *
+
+model_impl = {
+    'melgan': melgan,
+}
+
+
+def model_with_cfg(name, pretrained):
+    cfg = ConfigParser()
+    current_dir = os.path.dirname(os.path.abspath(__file__))
+    config_path = os.path.join(current_dir, '..', 'cfg', name + '.ini')
+    assert os.path.exists(config_path)
+    cfg.read(config_path)
+    arch = cfg.get('MODEL', 'ARCH')
+    model = model_impl[arch](cfg)
+    if pretrained:
+        checkpoint = cfg.get('MODEL', 'PRETRAINED_URL')
+        state_dict = hub.load_state_dict_from_url(checkpoint, progress=True, map_location='cpu')
+        model.load_state_dict(state_dict, strict=True)
+    return model, cfg
+
+
+def quant_melgan_8b(pretrained=True):
+    model, _ = model_with_cfg('quant_melgan_8b', pretrained)
+    return model
diff --git a/examples/text_to_speech/MelGAN/common.py b/brevitas_examples/text_to_speech/melgan/common.py
similarity index 100%
rename from examples/text_to_speech/MelGAN/common.py
rename to brevitas_examples/text_to_speech/melgan/common.py
diff --git a/examples/text_to_speech/MelGAN/generator_brevitas.py b/brevitas_examples/text_to_speech/melgan/generator_brevitas.py
similarity index 100%
rename from examples/text_to_speech/MelGAN/generator_brevitas.py
rename to brevitas_examples/text_to_speech/melgan/generator_brevitas.py
diff --git a/examples/text_to_speech/MelGAN/melgan.py b/brevitas_examples/text_to_speech/melgan/melgan.py
similarity index 100%
rename from examples/text_to_speech/MelGAN/melgan.py
rename to brevitas_examples/text_to_speech/melgan/melgan.py
diff --git a/examples/text_to_speech/MelGAN/res_stack_brevitas.py b/brevitas_examples/text_to_speech/melgan/res_stack_brevitas.py
similarity index 100%
rename from examples/text_to_speech/MelGAN/res_stack_brevitas.py
rename to brevitas_examples/text_to_speech/melgan/res_stack_brevitas.py
diff --git a/examples/text_to_speech/melgan_val.py b/brevitas_examples/text_to_speech/melgan_val.py
similarity index 66%
rename from examples/text_to_speech/melgan_val.py
rename to brevitas_examples/text_to_speech/melgan_val.py
index c55b71156..02d1154f7 100644
--- a/examples/text_to_speech/melgan_val.py
+++ b/brevitas_examples/text_to_speech/melgan_val.py
@@ -4,25 +4,24 @@
 import argparse
 from scipy.io.wavfile import write
 
-from MelGAN import *
 import torch.backends.cudnn as cudnn
 import brevitas.config
+from .MelGAN import model_with_cfg
 
 brevitas.config.IGNORE_MISSING_KEYS = False
 MAX_WAV_VALUE = 32768.0
-import configparser
 import random
 import os
 
 SEED = 123456
-models = {'melgan': melgan}
 
 parser = argparse.ArgumentParser()
 parser.add_argument('--input-folder', help='path to folder containing the val folder')
-parser.add_argument('--model-cfg', type=str, help='Path to pretrained model .ini configuration file')
 parser.add_argument('--workers', default=32, type=int, help='number of data loading workers')
 parser.add_argument('--batch-size', default=16, type=int, help='Minibatch size')
 parser.add_argument('--gpu', default=None, type=int, help='GPU id to use.')
+parser.add_argument('--pretrained', action='store_true', default=True, help='Load pretrained checkpoint')
+parser.add_argument('--model', type=str, default='quant_melgan_8b', help='Name of the model')
 
 
 def main():
@@ -30,30 +29,19 @@ def main():
     random.seed(SEED)
     torch.manual_seed(SEED)
 
-    assert os.path.exists(args.model_cfg)
-    cfg = configparser.ConfigParser()
-    cfg.read(args.model_cfg)
-    sampling_rate = cfg.getint('AUDIO', 'sampling_rate')
-
-    arch = cfg.get('MODEL', 'ARCH')
+    model, cfg = model_with_cfg(args.model, args.pretrained)
 
-    model = models[arch](cfg)
+    sampling_rate = cfg.getint('AUDIO', 'sampling_rate')
 
     if args.gpu is not None:
+        loc = 'cuda:{}'.format(args.gpu)
         torch.cuda.set_device(args.gpu)
         model = model.cuda(args.gpu)
         cudnn.benchmark = True
-    pretrained_url = cfg.get('MODEL', 'PRETRAINED_URL')
-    print("=> Loading checkpoint from:'{}'".format(pretrained_url))
-    if args.gpu is None:
-        loc = 'cpu'
-        checkpoint = torch.hub.load_state_dict_from_url(pretrained_url)
     else:
-        # Map model to be loaded to specified single gpu.
-        loc = 'cuda:{}'.format(args.gpu)
-        checkpoint = torch.hub.load_state_dict_from_url(pretrained_url, map_location=loc)
+        loc = 'cpu'
 
-    model.load_state_dict(checkpoint, strict=True)
+    model.to(loc)
     model.eval(inference=True)
 
     with torch.no_grad():
diff --git a/examples/text_to_speech/preprocess_dataset.py b/brevitas_examples/text_to_speech/preprocess_dataset.py
similarity index 76%
rename from examples/text_to_speech/preprocess_dataset.py
rename to brevitas_examples/text_to_speech/preprocess_dataset.py
index df76caa18..e2204efe5 100644
--- a/examples/text_to_speech/preprocess_dataset.py
+++ b/brevitas_examples/text_to_speech/preprocess_dataset.py
@@ -4,13 +4,13 @@
 import torch
 import argparse
 import numpy as np
-import configparser
+from configparser import ConfigParser
 
-from utilities.stft import TacotronSTFT
-from utilities.audio_processing import read_wav_np
+from .utilities.stft import TacotronSTFT
+from .utilities.audio_processing import read_wav_np
 
 
-def main(cfg, args):
+def preprocess(cfg, args):
     filter_length = cfg.getint('AUDIO', 'filter_length')
     hop_length = cfg.getint('AUDIO', 'hop_length')
     win_length = cfg.getint('AUDIO', 'win_length')
@@ -48,16 +48,21 @@ def main(cfg, args):
         torch.save(mel, melpath)
 
 
-if __name__ == '__main__':
+def main():
     parser = argparse.ArgumentParser()
-    parser.add_argument('-c', '--config', type=str, required=True,
-                        help="yaml file for config.")
+    parser.add_argument('-n', '--name', type=str, required=True,
+                        help="name of the model")
     parser.add_argument('-d', '--data-path', type=str, required=True,
                         help="root directory of wav files")
     args = parser.parse_args()
+    cfg = ConfigParser()
+    current_dir = os.path.dirname(os.path.abspath(__file__))
+    config_path = os.path.join(current_dir, 'cfg', args.name + '.ini')
+    assert os.path.exists(config_path)
+    cfg.read(config_path)
+
+    preprocess(cfg, args)
 
-    assert os.path.exists(args.config)
-    cfg = configparser.ConfigParser()
-    cfg.read(args.config)
 
-    main(cfg, args)
+if __name__ == '__main__':
+    main()
diff --git a/examples/text_to_speech/utilities/__init__.py b/brevitas_examples/text_to_speech/utilities/__init__.py
similarity index 100%
rename from examples/text_to_speech/utilities/__init__.py
rename to brevitas_examples/text_to_speech/utilities/__init__.py
diff --git a/examples/text_to_speech/utilities/audio_processing.py b/brevitas_examples/text_to_speech/utilities/audio_processing.py
similarity index 100%
rename from examples/text_to_speech/utilities/audio_processing.py
rename to brevitas_examples/text_to_speech/utilities/audio_processing.py
diff --git a/examples/text_to_speech/utilities/stft.py b/brevitas_examples/text_to_speech/utilities/stft.py
similarity index 100%
rename from examples/text_to_speech/utilities/stft.py
rename to brevitas_examples/text_to_speech/utilities/stft.py
diff --git a/examples/imagenet_classification/models/__init__.py b/examples/imagenet_classification/models/__init__.py
deleted file mode 100644
index 38de5032b..000000000
--- a/examples/imagenet_classification/models/__init__.py
+++ /dev/null
@@ -1,3 +0,0 @@
-from .mobilenetv1 import *
-from .vgg import *
-from .proxylessnas import *
\ No newline at end of file
diff --git a/examples/speech_to_text/quartznet/__init__.py b/examples/speech_to_text/quartznet/__init__.py
deleted file mode 100644
index 27976ead0..000000000
--- a/examples/speech_to_text/quartznet/__init__.py
+++ /dev/null
@@ -1,31 +0,0 @@
-# Adapted from https://github.com/NVIDIA/NeMo/blob/r0.9/collections/nemo_asr/
-# Copyright (C) 2020 Xilinx (Giuseppe Franco)
-# Copyright (C) 2019 NVIDIA CORPORATION.
-#
-# 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.
-
-
-# from .audio_preprocessing import AudioToMelSpectrogramPreprocessor
-from .data_layer import (
-        AudioToTextDataLayer)
-from .greedy_ctc_decoder import GreedyCTCDecoder
-from .quartznet import quartznet
-from .losses import CTCLossNM
-
-__all__ = ['AudioToTextDataLayer',
-           'quartznet']
-
-
-name = "quarznet_release"
diff --git a/examples/speech_to_text/requirements.txt b/examples/speech_to_text/requirements.txt
deleted file mode 100644
index b798cee7e..000000000
--- a/examples/speech_to_text/requirements.txt
+++ /dev/null
@@ -1,34 +0,0 @@
-pillow>=4.3.0
-ipython[all]
-tqdm
-sox
-ruamel.yaml
-jupyterlab
-tqdm
-boto3
-requests
-six
-ipdb
-h5py
-html2text
-nltk
-progressbar
-matplotlib
-wget 
-tensorboardX
-pandas
-onnx
-wget
-num2words
-librosa
-inflect
-kaldi-io
-marshmallow
-unidecode
-sentencepiece
-boto3
-matplotlib
-h5py
-youtokentome
-docrep
-torch-stft
diff --git a/noxfile.py b/noxfile.py
index db67bed70..6b5c88d88 100644
--- a/noxfile.py
+++ b/noxfile.py
@@ -10,61 +10,79 @@
 
 
 PYTORCH_CPU_VIRTUAL_PKG = '1.2.0'
+NOX_WIN_NUMPY_VERSION = '1.17.4'  # avoid errors from more recent Numpy called through Nox on Windows
 CONDA_PYTHON_IDS = tuple([f'conda_python_{i}' for i in CONDA_PYTHON_VERSIONS])
 PYTORCH_IDS = tuple([f'pytorch_{i}' for i in PYTORCH_VERSIONS])
 JIT_IDS = tuple([f'{i}'.lower() for i in JIT_STATUSES])
 
 
+def install_pytorch(pytorch, session):
+    is_win = system() == 'Windows'
+    is_cpu_virtual = version.parse(pytorch) >= version.parse(PYTORCH_CPU_VIRTUAL_PKG)
+    if is_cpu_virtual and is_win:
+        session.conda_install('-c', 'pytorch', f'pytorch=={pytorch}', f'numpy=={NOX_WIN_NUMPY_VERSION}', 'cpuonly')
+    elif is_cpu_virtual and not is_win:
+        session.conda_install('-c', 'pytorch', f'pytorch=={pytorch}', 'cpuonly')
+    elif not is_cpu_virtual and is_win:
+        session.conda_install('-c', 'pytorch', f'pytorch-cpu=={pytorch}', f'numpy=={NOX_WIN_NUMPY_VERSION}')
+    else:
+        session.conda_install('-c', 'pytorch', f'pytorch-cpu=={pytorch}')
+
+
+def dry_run_install_pytorch_deps(python, pytorch, session, deps_only):
+    deps = '--only-deps' if deps_only else '--no-deps'
+    is_win = system() == 'Windows'
+    is_cpu_virtual = version.parse(pytorch) >= version.parse(PYTORCH_CPU_VIRTUAL_PKG)
+    if is_cpu_virtual and is_win:
+        session.run('conda', 'create', '-n', 'dry_run', deps, '-d', '-c', 'pytorch', f'pytorch=={pytorch}',
+                    f'numpy=={NOX_WIN_NUMPY_VERSION}', 'cpuonly', f'python={python}')
+    elif is_cpu_virtual and not is_win:
+        session.run('conda', 'create', '-n', 'dry_run', deps, '-d', '-c', 'pytorch', f'pytorch=={pytorch}',
+                    'cpuonly', f'python={python}')
+    elif not is_cpu_virtual and not is_win:
+        session.run('conda', 'create', '-n', 'dry_run', deps, '-d', '-c', 'pytorch', f'pytorch-cpu=={pytorch}',
+                    'cpuonly', f'python={python}')
+    else:
+        session.run('conda', 'create', '-n', 'dry_run', deps, '-d', '-c', 'pytorch', f'pytorch-cpu=={pytorch}',
+                    f'numpy=={NOX_WIN_NUMPY_VERSION}', f'python={python}')
+
+
 @nox.session(venv_backend="conda", python=CONDA_PYTHON_VERSIONS)
 @nox.parametrize("pytorch", PYTORCH_VERSIONS, ids=PYTORCH_IDS)
 @nox.parametrize("jit_status", JIT_STATUSES, ids=JIT_IDS)
-def tests_cpu(session, pytorch, jit_status):
+def tests_brevitas_cpu(session, pytorch, jit_status):
     session.env['PYTORCH_JIT'] = '{}'.format(int(jit_status == 'jit_enabled'))
-    if version.parse(pytorch) >= version.parse(PYTORCH_CPU_VIRTUAL_PKG):
-        session.conda_install('-c', 'pytorch', f'pytorch=={pytorch}', 'cpuonly')
-    else:
-        session.conda_install('-c', 'pytorch', f'pytorch-cpu=={pytorch}')
+    install_pytorch(pytorch, session)
     session.install('.[test]')
-    session.run('pytest', '-v')
+    session.run('pytest', 'test/brevitas', '-v')
 
 
 @nox.session(venv_backend="conda", python=CONDA_PYTHON_VERSIONS)
 @nox.parametrize("pytorch", PYTORCH_VERSIONS, ids=PYTORCH_IDS)
-def tests_install_develop(session, pytorch):
-    if version.parse(pytorch) >= version.parse(PYTORCH_CPU_VIRTUAL_PKG):
-        session.conda_install('-c', 'pytorch', f'pytorch=={pytorch}', 'cpuonly')
-    else:
-        session.conda_install('-c', 'pytorch', f'pytorch-cpu=={pytorch}')
-    session.install('-e', '.')
-    if system() == 'Windows':
-        env = session.env
-        nox.command.run(['python', '-c', 'import brevitas'], env=env, path=os.path.dirname(session.bin))
-        nox.command.run(['python', '-c', 'import brevitas.function.ops_ste'], env=env,
-                        path=os.path.dirname(session.bin))
-    else:
-        session.run('python', '-c', 'import brevitas')
-        session.run('python', '-c', 'import brevitas.function.ops_ste')
+def tests_brevitas_install_dev(session, pytorch):
+    install_pytorch(pytorch, session)
+    session.install('-e', '.[test]')
+    session.run('pytest', '-v', 'test/brevitas/test_import.py')
+
+
+@nox.session(venv_backend="conda", python=CONDA_PYTHON_VERSIONS)
+@nox.parametrize("pytorch", PYTORCH_VERSIONS, ids=PYTORCH_IDS)
+def tests_brevitas_examples_install_dev(session, pytorch):
+    install_pytorch(pytorch, session)
+    session.conda_install('scipy')  # For Hadamard example
+    session.install('-e', '.[test, tts, stt]')
+    session.run('pytest', '-v', 'test/brevitas_examples/test_import.py')
 
 
 @nox.session(python=False)
 @nox.parametrize("pytorch", PYTORCH_VERSIONS, ids=PYTORCH_IDS)
 @nox.parametrize("python", CONDA_PYTHON_VERSIONS, ids=CONDA_PYTHON_IDS)
 def dry_run_pytorch_only_deps(session, pytorch, python):
-    if version.parse(pytorch) >= version.parse(PYTORCH_CPU_VIRTUAL_PKG):
-        session.run('conda', 'create', '-n', 'dry_run', '--only-deps', '-d', '-c', 'pytorch', f'pytorch=={pytorch}',
-                    'cpuonly', f'python={python}')
-    else:
-        session.run('conda', 'create', '-n', 'dry_run', '--only-deps', '-d', '-c', 'pytorch', f'pytorch-cpu=={pytorch}',
-                    f'python={python}')
+    dry_run_install_pytorch_deps(python, pytorch, session, deps_only=True)
 
 
 @nox.session(python=False)
 @nox.parametrize("pytorch", PYTORCH_VERSIONS, ids=PYTORCH_IDS)
 @nox.parametrize("python", CONDA_PYTHON_VERSIONS, ids=CONDA_PYTHON_IDS)
 def dry_run_pytorch_no_deps(session, pytorch, python):
-    if version.parse(pytorch) >= version.parse(PYTORCH_CPU_VIRTUAL_PKG):
-        session.run('conda', 'create', '-n', 'dry_run', '--no-deps', '-d', '-c', 'pytorch', f'pytorch=={pytorch}',
-                    'cpuonly', f'python={python}')
-    else:
-        session.run('conda', 'create', '-n', 'dry_run', '--no-deps', '-d', '-c', 'pytorch', f'pytorch-cpu=={pytorch}',
-                    f'python={python}')
+    dry_run_install_pytorch_deps(python, pytorch, session, deps_only=False)
diff --git a/requirements/requirements-stt.txt b/requirements/requirements-stt.txt
new file mode 100644
index 000000000..c820e73c5
--- /dev/null
+++ b/requirements/requirements-stt.txt
@@ -0,0 +1,10 @@
+pillow>=4.3.0
+ruamel.yaml
+requests
+librosa
+inflect
+unidecode
+torch-stft
+sox
+tqdm
+soundfile
diff --git a/examples/text_to_speech/requirements.txt b/requirements/requirements-tts.txt
similarity index 55%
rename from examples/text_to_speech/requirements.txt
rename to requirements/requirements-tts.txt
index f14b66145..075e3e6a0 100644
--- a/examples/text_to_speech/requirements.txt
+++ b/requirements/requirements-tts.txt
@@ -1,11 +1,7 @@
+soundfile
 librosa
-matplotlib
 numpy
 scipy
-tensorboardX
-torch
 tqdm
 pillow
 pyyaml
-soundfile
-packaging
diff --git a/setup.py b/setup.py
index 797e53dc6..567c0877e 100644
--- a/setup.py
+++ b/setup.py
@@ -167,15 +167,29 @@ def run(self):
       install_requires=read_requirements('requirements.txt'),
       extras_require={
           "Hadamard": read_requirements('requirements-hadamard.txt'),
-          "test": read_requirements('requirements-test.txt')
+          "test": read_requirements('requirements-test.txt'),
+          "tts": read_requirements('requirements-tts.txt'),
+          "stt": read_requirements('requirements-stt.txt')
       },
       packages=find_packages(),
       zip_safe=False,
       ext_modules=get_jittable_extension(),
       cmdclass={
-            'build_py': BuildPy,
-            'build_ext': BuildJittableExtension.with_options(no_python_abi_suffix=True),
-            'develop': DevelopInstall,
-        }
-      )
+          'build_py': BuildPy,
+          'build_ext': BuildJittableExtension.with_options(no_python_abi_suffix=True),
+          'develop': DevelopInstall,
+      },
+      package_data={
+          'brevitas_examples': ['*.ini', '*.yaml'],
+      },
+      entry_points={
+          'console_scripts': [
+              'brevitas_imagenet_val = brevitas_examples.imagenet_classification.imagenet_val:main',
+              'brevitas_quartznet_val = brevitas_examples.speech_to_text.quartznet_val:main',
+              'brevitas_melgan_val = brevitas_examples.text_to_speech.melgan_val:main',
+              'brevitas_quartznet_preprocess = brevitas_examples.speech_to_text.get_librispeech_data:main',
+              'brevitas_melgan_preprocess = brevitas_examples.text_to_speech.preprocess_dataset:main'
+          ],
+      })
+
 
diff --git a/test/common.py b/test/brevitas/common.py
similarity index 100%
rename from test/common.py
rename to test/brevitas/common.py
diff --git a/test/generate_quant_input.py b/test/brevitas/generate_quant_input.py
similarity index 100%
rename from test/generate_quant_input.py
rename to test/brevitas/generate_quant_input.py
diff --git a/test/test_act_scaling.py b/test/brevitas/test_act_scaling.py
similarity index 100%
rename from test/test_act_scaling.py
rename to test/brevitas/test_act_scaling.py
diff --git a/test/test_conv1d.py b/test/brevitas/test_conv1d.py
similarity index 100%
rename from test/test_conv1d.py
rename to test/brevitas/test_conv1d.py
diff --git a/test/brevitas/test_import.py b/test/brevitas/test_import.py
new file mode 100644
index 000000000..30a571e87
--- /dev/null
+++ b/test/brevitas/test_import.py
@@ -0,0 +1,44 @@
+# Copyright (c) 2019-     Xilinx, Inc              (Alessandro Pappalardo)
+# Copyright (c) 2016-     Facebook, Inc            (Adam Paszke)
+# Copyright (c) 2014-     Facebook, Inc            (Soumith Chintala)
+# Copyright (c) 2011-2014 Idiap Research Institute (Ronan Collobert)
+# Copyright (c) 2012-2014 Deepmind Technologies    (Koray Kavukcuoglu)
+# Copyright (c) 2011-2012 NEC Laboratories America (Koray Kavukcuoglu)
+# Copyright (c) 2011-2013 NYU                      (Clement Farabet)
+# Copyright (c) 2006-2010 NEC Laboratories America (Ronan Collobert, Leon Bottou, Iain Melvin, Jason Weston)
+# Copyright (c) 2006      Idiap Research Institute (Samy Bengio)
+# Copyright (c) 2001-2004 Idiap Research Institute (Ronan Collobert, Samy Bengio, Johnny Mariethoz)
+
+# All rights reserved.
+
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+
+# 1. Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+
+# 2. Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+
+# 3. Neither the names of Xilinx, Facebook, Deepmind Technologies, NYU,
+#    NEC Laboratories America and IDIAP Research Institute nor the names
+#    of its contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+# POSSIBILITY OF SUCH DAMAGE.
+
+
+def test_import_brevitas():
+    import brevitas
+    from brevitas.function import ops_ste
diff --git a/test/test_ops.py b/test/brevitas/test_ops.py
similarity index 100%
rename from test/test_ops.py
rename to test/brevitas/test_ops.py
diff --git a/test/test_quant.py b/test/brevitas/test_quant.py
similarity index 100%
rename from test/test_quant.py
rename to test/brevitas/test_quant.py
diff --git a/test/test_transposed_conv1d.py b/test/brevitas/test_transposed_conv1d.py
similarity index 100%
rename from test/test_transposed_conv1d.py
rename to test/brevitas/test_transposed_conv1d.py
diff --git a/test/brevitas_examples/test_import.py b/test/brevitas_examples/test_import.py
new file mode 100644
index 000000000..2b61a7b45
--- /dev/null
+++ b/test/brevitas_examples/test_import.py
@@ -0,0 +1,59 @@
+# Copyright (c) 2019-     Xilinx, Inc              (Alessandro Pappalardo)
+# Copyright (c) 2016-     Facebook, Inc            (Adam Paszke)
+# Copyright (c) 2014-     Facebook, Inc            (Soumith Chintala)
+# Copyright (c) 2011-2014 Idiap Research Institute (Ronan Collobert)
+# Copyright (c) 2012-2014 Deepmind Technologies    (Koray Kavukcuoglu)
+# Copyright (c) 2011-2012 NEC Laboratories America (Koray Kavukcuoglu)
+# Copyright (c) 2011-2013 NYU                      (Clement Farabet)
+# Copyright (c) 2006-2010 NEC Laboratories America (Ronan Collobert, Leon Bottou, Iain Melvin, Jason Weston)
+# Copyright (c) 2006      Idiap Research Institute (Samy Bengio)
+# Copyright (c) 2001-2004 Idiap Research Institute (Ronan Collobert, Samy Bengio, Johnny Mariethoz)
+
+# All rights reserved.
+
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+
+# 1. Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+
+# 2. Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+
+# 3. Neither the names of Xilinx, Facebook, Deepmind Technologies, NYU,
+#    NEC Laboratories America and IDIAP Research Institute nor the names
+#    of its contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+# POSSIBILITY OF SUCH DAMAGE.
+
+
+def test_import_image_classification():
+    from brevitas_examples.imagenet_classification import (
+        quant_mobilenet_v1_4b,
+        quant_proxylessnas_mobile14_hadamard_4b,
+        quant_proxylessnas_mobile14_4b5b,
+        quant_proxylessnas_mobile14_4b)
+
+
+def test_import_tts():
+    from brevitas_examples.text_to_speech import quant_melgan_8b
+
+
+def test_import_stt():
+    from brevitas_examples.speech_to_text import (
+        quant_quartznet_pertensorscaling_8b,
+        quant_quartznet_perchannelscaling_8b,
+        quant_quartznet_perchannelscaling_4b
+    )
\ No newline at end of file