quick_test.py

"""
Quick LoRA Test - Works without DLL issues
Uses llama.cpp Python bindings
"""
import sys
import os

# Add llama.cpp to path
sys.path.insert(0, '../../llama.cpp/build/bin')

try:
    from llama_cpp import Llama
    print("✅ llama-cpp-python available")
except ImportError:
    print("❌ llama-cpp-python not installed")
    print("\nInstall it with:")
    print('  pip install llama-cpp-python')
    print('\nOr test via API after starting the server without --lora')
    sys.exit(1)

# Test configuration
BASE_MODEL = "../../SmolLM2.Q4_K_M.gguf"
LORA_PATH = "./lora-output/Lora-Output-F16-LoRA.gguf"
PROMPT = "Once upon a time in a magical forest"

print("\n" + "="*70)
print("Testing Base Model vs LoRA-Enhanced Model")
print("="*70)

# Test 1: Without LoRA
print("\n🔵 Test 1: BASE MODEL (no LoRA)")
print("-"*70)
try:
    llm = Llama(model_path=BASE_MODEL, n_ctx=512, n_threads=4, verbose=False)
    result = llm(PROMPT, max_tokens=60, temperature=0.7, stop=["\n\n"])
    print(f"Prompt: {PROMPT}")
    print(f"Output: {result['choices'][0]['text']}")
except Exception as e:
    print(f"❌ Error: {e}")

# Test 2: With LoRA  
print("\n🟢 Test 2: WITH LORA ADAPTER")
print("-"*70)
try:
    llm_lora = Llama(model_path=BASE_MODEL, lora_path=LORA_PATH, n_ctx=512, n_threads=4, verbose=False)
    result = llm_lora(PROMPT, max_tokens=60, temperature=0.7, stop=["\n\n"])
    print(f"Prompt: {PROMPT}")
    print(f"Output: {result['choices'][0]['text']}")
    print("\n✅ LoRA is working!")
except Exception as e:
    print(f"❌ Error: {e}")

print("\n" + "="*70)