[TVMRT] Document missing dependency for tvm python package on 11.2 SDK

11.2 SDK is missing psutil and typing_extensions python package
which is a dependency for tvm python wheel.

Signed-off-by: Abhay-Chirania <a-chirania@ti.com>

Author: abhaychirania2411

README.md

@@ -401,7 +401,11 @@ You have two options for setting up the EdgeAI TIDL Tools repository on your TI
    python3 basic_example.py --config ./config.yaml -r tflitert --infer
 
    # Running tvmrt examples
-   python3 basic_example.py --config ./config.yaml -r tvmrt --infer
+   # Note: The TVM python wheel requires additional dependencies that are 
+   # currently not provided in SDK 11.2. Please run
+   # `pip install psutil typing_extensions` on SoC to enable tvm inference
+   # 
+   #python3 basic_example.py --config ./config.yaml -r tvmrt --infer
 
    cd -
    ```

runtimes/tidl_wrapper/python/tvmrt/README.md

@@ -23,6 +23,13 @@ This module is designed to simplify the process of using TVM Runtime with TIDL a
 - NumPy
 - TIDL Tools (for model compilation)
 
+## Note
+
+For inference on target machine, user needs to manually install additional dependencies using the below command:
+``` python
+pip3 install psutil typing_extensions
+```
+
 ## API Documentation
 
 ### TVMRT Class
@@ -126,6 +133,7 @@ def get_performance(self):
               'subgraph_time': (value, "ms") - Total TIDL Subgraphs processing time
               'read_total': (value, "bytes") - Total DDR Read bytes [X for x86 runs]
               'write_total': (value, "bytes") - Total DDR Write bytes [X for x86 runs]
+              'num_subgraphs': (value, "") - Total Detected subgraphs
     """
 ```
 
@@ -270,6 +278,7 @@ The `get_performance()` method returns a dictionary with the following metrics,
 | `subgraph_time` | Time spent in all TIDL subgraph execution | milliseconds (ms) |
 | `read_total` | Total DDR read bytes | bytes |
 | `write_total` | Total DDR write bytes | bytes |
+| `num_subgraphs` | Total Subgraphs detected | |
 
 ## Model Information
 

runtimes/tidl_wrapper/python/tvmrt/tvmrt_wrapper.py

@@ -2,11 +2,28 @@
 import sys
 import numpy as np
 import platform
+import platform
 
 # set the environment variable before importing TVM so that
 # it takes effect while loading and initializing the c++ library.
 os.environ["TIDL_RT_PERFSTATS"] = "1"
-import tvm
+
+try:
+    import tvm
+except ModuleNotFoundError as e:
+    print("\n" + "="*80)
+    print("ERROR: TVMRT Dependencies Missing")
+    print("="*80)
+    if platform.machine() != "aarch64":
+        print("Please follow the setup steps before running the application.")
+    else:
+        print(f"Failed to import required module: {e.name if hasattr(e, 'name') else 'tvm dependencies'}")
+        print("\nTVMRT requires additional Python packages to function properly.")
+        print("Please install the missing dependencies using the following command:")
+        print("\n  pip3 install psutil typing_extensions")
+        print("\nAfter installation, please re-run your application.")
+    print("="*80 + "\n")
+    sys.exit(1)
 
 class TVMRT:
     """
@@ -131,6 +148,7 @@ def get_performance(self):
         'subgraph_time': Total TIDL Subgraphs processing time (ms)
         'read_total': Total DDR Read bytes [X for x86 runs]
         'write_total': Total DDR Write bytes [X for x86 runs]
+        'num_subgraphs': Total Detected subgraphs
 
         Returns:
             dict: performance_name : (performance_value, unit)
@@ -169,7 +187,8 @@ def get_performance(self):
                  'core_time':       (core_time,"ms"),
                  'subgraph_time':   (subgraph_time,"ms"),
                  'read_total':      (read_total, "bytes"),
-                 'write_total':     (write_total, "bytes")
+                 'write_total':     (write_total, "bytes"),
+                 'num_subgraphs':   (len(subgraphIds), "")
                 }
 
         return stats