test_dwrko.py

#!/usr/bin/env python3
"""
Dwrko-M1.0 Testing Script
Test your fine-tuned Claude-like AI assistant
"""

import torch
import argparse
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import time

def load_dwrko_model(model_path):
    """Load fine-tuned Dwrko-M1.0 model"""
    
    print(f"🤖 Loading Dwrko-M1.0 from {model_path}")
    
    # Load base tokenizer
    tokenizer = AutoTokenizer.from_pretrained("bigcode/starcoder2-3b")
    if tokenizer.pad_token is None:
        tokenizer.pad_token = tokenizer.eos_token
    
    # Load base model
    base_model = AutoModelForCausalLM.from_pretrained(
        "bigcode/starcoder2-3b",
        torch_dtype=torch.float16,
        device_map="auto"
    )
    
    # Load LoRA adapters
    model = PeftModel.from_pretrained(base_model, model_path)
    model = model.merge_and_unload()  # Merge adapters for faster inference
    
    print("✅ Dwrko-M1.0 loaded successfully!")
    return model, tokenizer

def generate_response(model, tokenizer, prompt, max_length=512, temperature=0.7):
    """Generate response from Dwrko-M1.0"""
    
    # Format prompt
    formatted_prompt = f"### Instruction:\n{prompt}\n\n### Response:\n"
    
    # Tokenize
    inputs = tokenizer(formatted_prompt, return_tensors="pt").to(model.device)
    
    # Generate
    start_time = time.time()
    with torch.no_grad():
        outputs = model.generate(
            inputs.input_ids,
            max_length=max_length,
            temperature=temperature,
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id,
            eos_token_id=tokenizer.eos_token_id,
            top_p=0.9,
            repetition_penalty=1.1
        )
    
    generation_time = time.time() - start_time
    
    # Decode response
    full_response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    response = full_response.split("### Response:\n")[-1].strip()
    
    # Calculate tokens per second
    output_tokens = len(outputs[0]) - len(inputs.input_ids[0])
    tokens_per_second = output_tokens / generation_time if generation_time > 0 else 0
    
    return response, tokens_per_second

def run_test_suite(model, tokenizer):
    """Run comprehensive test suite for Dwrko-M1.0"""
    
    print("\n" + "="*60)
    print("🧪 Running Dwrko-M1.0 Test Suite")
    print("="*60)
    
    test_prompts = [
        # Coding Tests
        {
            "category": "💻 Coding",
            "prompt": "Write a Python function to calculate the factorial of a number using recursion.",
            "expected_keywords": ["def", "factorial", "return", "if", "else"]
        },
        {
            "category": "💻 Coding", 
            "prompt": "How do you reverse a string in Python? Show me 3 different methods.",
            "expected_keywords": ["[::-1]", "reversed", "for", "range"]
        },
        {
            "category": "💻 Coding",
            "prompt": "Write a function to check if a number is prime.",
            "expected_keywords": ["def", "prime", "for", "range", "return"]
        },
        
        # Reasoning Tests
        {
            "category": "🧠 Reasoning",
            "prompt": "If a train travels 120 miles in 2 hours, what is its average speed?",
            "expected_keywords": ["60", "mph", "speed", "miles", "hour"]
        },
        {
            "category": "🧠 Reasoning",
            "prompt": "Solve this equation: 2x + 5 = 13. Show your work.",
            "expected_keywords": ["x", "4", "subtract", "divide", "2x"]
        },
        {
            "category": "🧠 Reasoning",
            "prompt": "What is the next number in this sequence: 2, 4, 8, 16, ?",
            "expected_keywords": ["32", "double", "multiply", "pattern"]
        },
        
        # Explanation Tests
        {
            "category": "📚 Explanation",
            "prompt": "Explain what machine learning is in simple terms.",
            "expected_keywords": ["algorithm", "data", "learn", "pattern", "computer"]
        },
        {
            "category": "📚 Explanation",
            "prompt": "What is the difference between a list and a tuple in Python?",
            "expected_keywords": ["mutable", "immutable", "[]", "()", "change"]
        }
    ]
    
    total_tests = len(test_prompts)
    passed_tests = 0
    total_tokens_per_second = 0
    
    for i, test in enumerate(test_prompts, 1):
        print(f"\n🔍 Test {i}/{total_tests} - {test['category']}")
        print(f"❓ Prompt: {test['prompt']}")
        
        # Generate response
        response, tps = generate_response(model, tokenizer, test['prompt'])
        
        print(f"🤖 Dwrko-M1.0: {response[:200]}{'...' if len(response) > 200 else ''}")
        print(f"⚡ Speed: {tps:.1f} tokens/second")
        
        # Check if response contains expected keywords
        response_lower = response.lower()
        found_keywords = sum(1 for keyword in test['expected_keywords'] 
                           if keyword.lower() in response_lower)
        
        if found_keywords >= len(test['expected_keywords']) // 2:  # At least half keywords found
            print("✅ Test PASSED")
            passed_tests += 1
        else:
            print("❌ Test FAILED")
            print(f"   Expected keywords: {test['expected_keywords']}")
        
        total_tokens_per_second += tps
        print("-" * 60)
    
    # Final results
    print(f"\n📊 Test Results Summary:")
    print(f"✅ Passed: {passed_tests}/{total_tests} ({passed_tests/total_tests*100:.1f}%)")
    print(f"⚡ Average Speed: {total_tokens_per_second/total_tests:.1f} tokens/second")
    
    if passed_tests/total_tests >= 0.7:
        print("🎉 Dwrko-M1.0 is performing well!")
    else:
        print("⚠️  Consider additional training or parameter tuning")

def interactive_mode(model, tokenizer):
    """Interactive chat with Dwrko-M1.0"""
    
    print("\n" + "="*60)
    print("💬 Interactive Mode - Chat with Dwrko-M1.0")
    print("Type 'quit' to exit")
    print("="*60)
    
    while True:
        user_input = input("\n👤 You: ").strip()
        
        if user_input.lower() in ['quit', 'exit', 'q']:
            print("👋 Goodbye!")
            break
        
        if not user_input:
            continue
        
        print("🤖 Dwrko-M1.0: ", end="", flush=True)
        response, tps = generate_response(model, tokenizer, user_input, max_length=256)
        print(response)
        print(f"   ⚡ {tps:.1f} tokens/sec")

def main():
    parser = argparse.ArgumentParser(description="Test Dwrko-M1.0 Model")
    parser.add_argument("--model_path", required=True, help="Path to fine-tuned Dwrko-M1.0")
    parser.add_argument("--test_suite", action="store_true", help="Run automated test suite")
    parser.add_argument("--interactive", action="store_true", help="Start interactive chat")
    parser.add_argument("--single_test", type=str, help="Test single prompt")
    
    args = parser.parse_args()
    
    # Load model
    model, tokenizer = load_dwrko_model(args.model_path)
    
    if args.test_suite:
        run_test_suite(model, tokenizer)
    
    if args.single_test:
        print(f"\n🔍 Testing single prompt: {args.single_test}")
        response, tps = generate_response(model, tokenizer, args.single_test)
        print(f"🤖 Dwrko-M1.0: {response}")
        print(f"⚡ Speed: {tps:.1f} tokens/second")
    
    if args.interactive:
        interactive_mode(model, tokenizer)
    
    if not any([args.test_suite, args.interactive, args.single_test]):
        print("\n⚠️  Please specify --test_suite, --interactive, or --single_test")
        print("Example: python test_dwrko.py --model_path ./dwrko-m1.0 --test_suite")

if __name__ == "__main__":
    main()