Fix M5 proofpoint

Fix the issue where the pretrained_models and datasets directory
was not being copied with pip install not using -e.

demo_helpers/MANIFEST.in

@@ -0,0 +1,3 @@
+include demo_helpers/datasets/README.md
+include demo_helpers/pretrained_models/m5.pt
+include demo_helpers/pretrained_models/pointnet.pth

demo_helpers/setup.py

@@ -10,6 +10,7 @@
     packages=find_packages(
         exclude=["*.__pycache__.*"],
     ),
+    include_package_data=True,
     install_requires=[
         "charset-normalizer==3.3.2",
         "transformers>=4.20.0",

proof_points/speech/m5/README.md

@@ -19,6 +19,19 @@ M5's Keyword Spotting accuracy is evaluated using the [SpeechCommands dataset](h
   pip install -r requirements.txt
   ```
 
+- Since this proofpoint uses audio files, often the audio libraries must be installed on system.
+  - For Ubuntu OS:
+
+  ```bash
+  sudo apt install libsox-dev
+  ```
+
+  - For Rocky OS:
+
+  ```bash
+  sudo dnf install sox-devel
+  ```
+
 ## Build and Evaluate
 
 To build and evaluate M5: