quick_start.py

"""
Quick start example - Minimal code to run Qwen ONNX model
"""

import onnxruntime_genai as og
from pathlib import Path

# Point to your ONNX model directory
MODEL_DIR = Path(__file__).parent

# Load model and tokenizer
print(f"Loading model from {MODEL_DIR}")
model = og.Model(str(MODEL_DIR))
tokenizer = og.Tokenizer(model)

# Simple text generation
prompt = "Hello, how are you?"
print(f"\nPrompt: {prompt}")

# Encode input
input_tokens = tokenizer.encode(prompt)

# Setup generation parameters
config = model.get_default_generation_search_parameters()
config.max_length = 100
config.temperature = 0.6

# Generate
generator = og.Generator(model, config)
generator.append_tokens(input_tokens)

while not generator.is_done():
    generator.compute_logits()
    generator.generate_next_token()

# Decode output
output_tokens = generator.get_sequence(0)
output_text = tokenizer.decode(output_tokens)

print(f"Response: {output_text}")