scripts/verify_l4_deployment.py

import requests
import json
import time

BASE_URL = "https://conquerorr000-turkish-medical-model-api.hf.space"

def verify_deployment():
    """Verify L4 optimization deployment"""
    print("� L4 Optimization Deployment Verification")
    print("=" * 50)
    
    # Check health
    try:
        response = requests.get(f"{BASE_URL}/health", timeout=30)
        health_data = response.json()
        
        if health_data.get("model_loaded"):
            print(" Model Status: Loaded and Ready")
        else:
            print(" Model Status: Not Ready")
            return False
            
    except Exception as e:
        print(f" Health Check Failed: {e}")
        return False
    
    # Check new endpoints
    endpoints_to_check = ["/memory-status", "/conversation"]
    
    for endpoint in endpoints_to_check:
        try:
            if endpoint == "/conversation":
                # Test with POST data
                test_data = {
                    "messages": [{"role": "user", "content": "Test"}],
                    "max_tokens": 50
                }
                response = requests.post(f"{BASE_URL}{endpoint}", json=test_data, timeout=30)
            else:
                response = requests.get(f"{BASE_URL}{endpoint}", timeout=30)
            
            if response.status_code == 200:
                print(f" Endpoint {endpoint}: Working")
            else:
                print(f" Endpoint {endpoint}: Failed ({response.status_code})")
                
        except Exception as e:
            print(f" Endpoint {endpoint}: Error - {e}")
    
    # Check debug info for L4 optimizations
    try:
        response = requests.get(f"{BASE_URL}/debug", timeout=30)
        debug_data = response.json()
        
        optimization_info = debug_data.get("optimization_info", {})
        
        print(f"\n L4 Optimization Status:")
        print(f"   Precision: {optimization_info.get('precision', 'Unknown')}")
        print(f"   Quantization: {optimization_info.get('quantization', 'Unknown')}")
        print(f"   Flash Attention: {optimization_info.get('flash_attention', 'Unknown')}")
        print(f"   TF32: {optimization_info.get('tf32', 'Unknown')}")
        print(f"   LoRA Merged: {optimization_info.get('lora_merged', 'Unknown')}")
        
        # Check if optimizations are active
        optimizations_active = (
            optimization_info.get('precision') == 'FP16' and
            optimization_info.get('quantization') == 'None' and
            optimization_info.get('flash_attention') == 'Enabled'
        )
        
        if optimizations_active:
            print(" L4 Optimizations: Active")
        else:
            print(" L4 Optimizations: May not be fully active")
            
    except Exception as e:
        print(f" Debug Check Failed: {e}")
    
    # Quick performance test
    try:
        print(f"\n Quick Performance Test:")
        start_time = time.time()
        
        response = requests.get(f"{BASE_URL}/test", timeout=60)
        
        if response.status_code == 200:
            test_data = response.json()
            generation_time = test_data.get("generation_time", 0)
            
            print(f"   Generation Time: {generation_time:.2f}s")
            
            if generation_time < 6:
                print(" Performance: Excellent")
            elif generation_time < 8:
                print(" Performance: Good")
            else:
                print(" Performance: Needs Optimization")
                
        else:
            print(" Performance Test Failed")
            
    except Exception as e:
        print(f" Performance Test Error: {e}")
    
    print(f"\n{'='*50}")
    print(" Verification Complete!")

if __name__ == "__main__":
    verify_deployment()