INNER CODE UNIT · Python
get_args
KodCode-AI/kodcode · pipeline/completion_open_model.py:18
def get_args():
# Experiment Settings
parser = argparse.ArgumentParser(description="Response Generation Manager.")
parser.add_argument("--model_path", type=str, default="Qwen/Qwen2.5-7B-Instruct",
help="We will support more models in the future.")
parser.add_argument("--input_file", type=str, default=None, help="Input dataset file name")
parser.add_argument("--batch_size", type=int, default=128, help="Number of samples per batch")
parser.add_argument("--checkpoint_every", type=int, default=20, help="Save checkpoint every n batches")
parser.add_argument("--api_url", type=str, default="https://api.together.xyz/v1/chat/completions", help="API URL")
parser.add_argument("--api_key", type=str, default="", help="Together API Key (start without Bearer)")
parser.add_argument("--offline", action="store_true", help="Use local engine")
# Generation Parameters
parser.add_argument('--engine', default="vllm", type=str, choices=["vllm", "hf", "together"])
parser.add_argument("--device", type=str, default="0")
parser.add_argument("--dtype", type=str, default="bfloat16", choices=["float16", "bfloat16"])
parser.add_argument("--tensor_parallel_size", type=int, default=1, help="Number of GPUs to use for tensor parallelism. Only used for Llama 70B models.")
parser.add_argument("--gpu_memory_utilization", type=float, default=0.95)