platform/aiml/experiments/test_model.py

#!/usr/bin/env python3
"""
Test script to verify the trained model works correctly.
"""

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
import time

def test_model_loading():
    """Test that the model loads successfully."""
    print("🧪 Testing model loading...")
    
    model_path = "/home/x/adaptai/experiments/qwen3-8b-elizabeth-simple/"
    
    # Load tokenizer
    print("📥 Loading tokenizer...")
    tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
    print("✅ Tokenizer loaded successfully!")
    
    # Load model
    print("📥 Loading model...")
    start_time = time.time()
    
    model = AutoModelForCausalLM.from_pretrained(
        model_path,
        torch_dtype=torch.bfloat16,
        device_map="auto",
        trust_remote_code=True
    )
    
    load_time = time.time() - start_time
    print(f"✅ Model loaded successfully in {load_time:.2f} seconds!")
    
    # Check device placement
    print(f"📊 Model device: {model.device}")
    print(f"📊 Model dtype: {model.dtype}")
    
    return model, tokenizer

def test_inference(model, tokenizer):
    """Test basic inference."""
    print("\n🧪 Testing inference...")
    
    # Test prompt
    prompt = "Hello, how are you today?"
    
    print(f"📝 Prompt: {prompt}")
    
    # Tokenize
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    print(f"🔢 Input tokens: {inputs.input_ids.shape}")
    
    # Generate
    start_time = time.time()
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=50,
            do_sample=True,
            temperature=0.7,
            pad_token_id=tokenizer.eos_token_id
        )
    
    gen_time = time.time() - start_time
    
    # Decode
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    
    print(f"✅ Generation completed in {gen_time:.2f} seconds!")
    print(f"💬 Response: {response}")
    
    return response

def test_tool_use_capability(model, tokenizer):
    """Test tool use capability."""
    print("\n🧪 Testing tool use capability...")
    
    tool_prompt = """Please help me calculate the square root of 144 using the calculator tool.

Available tools:
- calculator: Performs mathematical calculations

Please respond with the tool call format."""
    
    print(f"📝 Tool prompt: {tool_prompt[:100]}...")
    
    inputs = tokenizer(tool_prompt, return_tensors="pt").to(model.device)
    
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=100,
            do_sample=True,
            temperature=0.3,
            pad_token_id=tokenizer.eos_token_id
        )
    
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    
    print(f"💬 Tool use response: {response}")
    
    # Check if response contains tool-like patterns
    if "calculator" in response.lower() or "tool" in response.lower():
        print("✅ Tool use capability detected!")
        return True
    else:
        print("⚠️  Tool use pattern not clearly detected")
        return False

if __name__ == "__main__":
    print("=" * 60)
    print("🤖 Qwen3-8B-Elizabeth-Simple Model Test")
    print("=" * 60)
    
    try:
        model, tokenizer = test_model_loading()
        test_inference(model, tokenizer)
        tool_use_detected = test_tool_use_capability(model, tokenizer)
        
        print("\n" + "=" * 60)
        print("🎉 Model Test Summary:")
        print(f"   ✅ Model loading: Successful")
        print(f"   ✅ Basic inference: Working")
        print(f"   ✅ Tool use capability: {'Detected' if tool_use_detected else 'Needs verification'}")
        print("   📍 Model is ready for deployment!")
        print("=" * 60)
        
    except Exception as e:
        print(f"\n❌ Test failed with error: {e}")
        raise