Move qwen3_omni_case_compare.py to tools/ and reorganize

- Move qwen3_omni_case_compare.py into samples/tools/
- Rename requirements.txt to requirements_hf_tokenizer_to_ov.txt
  (per-tool naming convention for independent dependencies)
- Update tools/README.md with both tools documented
- Update bat launcher and modeling README with new paths
- qwen3_omni_case_compare.py uses stdlib only, no extra requirements

Signed-off-by: Zhang, Xiaolin <xiaolin.zhang@intel.com>

Author: xzhan34

src/cpp/src/modeling/README.md

@@ -21,8 +21,10 @@ modeling/
 ├── samples/              # Sample executables
 │   ├── modeling_qwen3_omni.cpp          # Case 1: image+text → text
 │   ├── modeling_qwen3_omni_tts_min.cpp  # Cases 2–5: multimodal → text + TTS
-│   └── extract_video_frames.cpp         # Video frame extraction tool
-└── qwen3_omni_case_compare.py           # Automated test harness (all cases × devices × precisions)
+│   ├── extract_video_frames.cpp         # Video frame extraction tool
+│   └── tools/                           # Dev-only Python utilities (see tools/README.md)
+│       └── qwen3_omni_case_compare.py   # Automated test harness (all cases × devices × precisions)
+│       └── hf_tokenizer_to_ov.py        # Convert a HuggingFace tokenizer to OpenVINO IR format.
 ```
 
 ## Prerequisites
@@ -199,13 +201,13 @@ Control inference precision and KV-cache compression via the `--precision` argum
 
 Aliases: `fp32_kv8` → `inf_fp32_kv_int8`, `fp16_kv8` → `inf_fp16_kv_int8`, etc.
 
-## Automated Case Comparison (`qwen3_omni_case_compare.py`)
+## Automated Case Comparison (`tools/qwen3_omni_case_compare.py`)
 
 Runs all cases across multiple devices and precision modes, generating a JSON report
 with performance metrics and text outputs for comparison.
 
 ```bat
-python qwen3_omni_case_compare.py ^
+python tools/qwen3_omni_case_compare.py ^
     --model-dir     D:\models\Qwen3-Omni-4B-Instruct-multilingual ^
     --image         path\to\image.jpg ^
     --test-audio    path\to\audio.wav ^

src/cpp/src/modeling/samples/tools/README.md

@@ -1,11 +1,13 @@
 # Dev Tools
 
-Dev-only Python utilities for model preparation. **Not included in release builds.**
+Dev-only Python utilities for model preparation and testing. **Not included in release builds.**
 
 ## Setup
 
+Each tool may have its own `requirements_<tool>.txt`. Install as needed:
+
 ```bash
-pip install -r requirements.txt
+pip install -r requirements_hf_tokenizer_to_ov.txt
 ```
 
 ## Tools
@@ -14,6 +16,8 @@ pip install -r requirements.txt
 
 Convert a HuggingFace tokenizer to OpenVINO IR (`openvino_tokenizer.xml` + `openvino_detokenizer.xml`).
 
+Dependencies: `requirements_hf_tokenizer_to_ov.txt`
+
 ```bash
 # Basic usage — output defaults to model_id directory
 python hf_tokenizer_to_ov.py <model_id_or_path>
@@ -27,3 +31,24 @@ Options:
 | Flag | Description |
 |------|-------------|
 | `-o, --output-dir` | Output directory (default: model_id directory) |
+
+---
+
+### qwen3_omni_case_compare.py
+
+Automated test harness: runs Qwen3-Omni inference cases across multiple devices and precision modes, generating a JSON report with performance metrics and text outputs.
+
+Dependencies: Python standard library only (no extra install needed).
+
+```bat
+python qwen3_omni_case_compare.py ^
+    --model-dir D:\models\Qwen3-Omni-4B-Instruct-multilingual ^
+    --image path\to\image.jpg ^
+    --test-audio path\to\audio.wav ^
+    --cpp-bin path\to\modeling_qwen3_omni.exe ^
+    --out-json reports\case_compare.json ^
+    --devices CPU,GPU ^
+    --precisions fp32,inf_fp16_kv_int8
+```
+
+See [modeling README](../../../modeling/README.md) for full options.