fix_voice_gpu.sh

#!/bin/bash
# 🎤 FIX VOICE GPU PROCESSING 🔊

echo "🚀 FIXING GPU VOICE PROCESSING..."

# Set CUDA paths
export CUDA_HOME=/usr/local/cuda
export PATH=$CUDA_HOME/bin:$PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH

# Fix cuDNN path
export LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATH

# Install ctranslate2 with CUDA support
echo "📦 Installing CTranslate2 with CUDA..."
pip install --upgrade --force-reinstall ctranslate2 --no-cache-dir

# Install faster-whisper with proper dependencies
echo "📦 Installing Faster-Whisper for GPU..."
pip install --upgrade --force-reinstall "faster-whisper>=1.0.0" --no-cache-dir

echo "✅ GPU voice processing fixed!"
echo ""
echo "Now updating golem_flask_server.py for GPU..."

# Fix the ASR initialization in golem_flask_server.py
python3 - <<'EOF'
import fileinput
import sys

import os
script_dir = os.path.dirname(os.path.abspath(__file__))
file_path = os.path.join(script_dir, "home", "chezy", "golem_flask_server.py")

# Read and fix the file
with open(file_path, 'r') as f:
    lines = f.readlines()

# Fix the ASR initialization
for i, line in enumerate(lines):
    # Force GPU for CT2
    if 'os.environ.setdefault("CT2_USE_CUDA"' in line:
        lines[i] = '    os.environ.setdefault("CT2_USE_CUDA", "1")  # FORCE GPU\n'
        print(f"✅ Fixed line {i+1}: Forced GPU for CT2")
    
    # Use int8_float16 for RTX 3050
    elif '"FASTER_WHISPER_COMPUTE_TYPE"' in line and 'getenv' in line:
        lines[i] = '    compute_type = os.getenv("FASTER_WHISPER_COMPUTE_TYPE", "int8_float16")  # RTX 3050 optimized\n'
        print(f"✅ Fixed line {i+1}: Set compute type for RTX 3050")
    
    # Force device to cuda in WhisperModel
    elif '_faster_whisper_model = WhisperModel(' in line:
        # Check if device parameter exists
        if 'device=' not in lines[i]:
            lines[i] = lines[i].rstrip()[:-1] + ', device="cuda")\n'
            print(f"✅ Fixed line {i+1}: Added device='cuda' to WhisperModel")

# Write back
with open(file_path, 'w') as f:
    f.writelines(lines)

print("✅ golem_flask_server.py updated for GPU!")
EOF

echo ""
echo "🎯 TESTING GPU VOICE..."
python3 - <<'EOF'
import os
os.environ["CT2_USE_CUDA"] = "1"
os.environ["CUDA_VISIBLE_DEVICES"] = "0"

try:
    from faster_whisper import WhisperModel
    print("✅ Faster-Whisper imported successfully!")
    
    # Try to load model on GPU with int8_float16
    model = WhisperModel(
        "Systran/faster-distil-whisper-large-v3",
        device="cuda",
        compute_type="int8_float16"
    )
    print("✅ Whisper model loaded on GPU!")
    print("🚀 GPU VOICE PROCESSING READY!")
except Exception as e:
    print(f"❌ Error: {e}")
    print("\nTrying fallback to float16...")
    try:
        model = WhisperModel(
            "Systran/faster-distil-whisper-large-v3",
            device="cuda", 
            compute_type="float16"
        )
        print("✅ Whisper model loaded on GPU with float16!")
    except Exception as e2:
        print(f"❌ Float16 also failed: {e2}")
EOF

echo ""
echo "✅ DONE! Restart the server with:"
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
ROOT_DIR="$(dirname "$SCRIPT_DIR")"
echo "   cd $ROOT_DIR && ./start_consciousness_ecosystem.sh"