Enhance batch_confirm_hf_optimized.py to ensure torchcodec is installed before loading the dataset, and update requirements_optimized.txt to include torchcodec. Modify run_optimized_192cores_no_root.sh to install additional audio dependencies and test audio imports.

vosk/test_files/batch_confirm_hf_optimized.py

@@ -27,6 +27,16 @@ BATCH_SIZE = 32  # Increased batch size for better throughput
 MAX_CONCURRENT_REQUESTS = 48  # 192/4 for optimal concurrency
 CHUNK_SIZE = 1000  # Process data in chunks to manage memory
 
+# Ensure torchcodec is installed before loading dataset
+try:
+    import torchcodec
+except ImportError:
+    print("Installing torchcodec...")
+    import subprocess
+    import sys
+    subprocess.check_call([sys.executable, "-m", "pip", "install", "torchcodec>=0.1.0"])
+    import torchcodec
+
 # Load the dataset with audio decoding
 print("Loading dataset...")
 ds = load_dataset(