linalg_zero/grpo/scripts/run.py

# Copyright Sierra

import argparse
import logging

from dotenv import load_dotenv
from litellm import provider_list

from linalg_zero.grpo.envs.user import UserStrategy
from linalg_zero.grpo.run import run
from linalg_zero.grpo.types import RunConfig

load_dotenv(override=True)

# Suppress LiteLLM logging spam
logging.getLogger("LiteLLM").setLevel(logging.WARNING)


def parse_args() -> RunConfig:
    parser = argparse.ArgumentParser()
    parser.add_argument("--project-id", type=int, default=0)
    parser.add_argument("--project-name", type=str, default="linalgzero-grpo-eval")
    parser.add_argument("--num-trials", type=int, default=1)
    parser.add_argument("--env", type=str, choices=["linear_algebra"], default="linear_algebra")
    parser.add_argument(
        "--model",
        type=str,
        help="The model to use for the agent",
    )
    parser.add_argument(
        "--model-provider",
        type=str,
        choices=provider_list,
        help="The model provider for the agent",
    )
    parser.add_argument(
        "--user-model",
        type=str,
        default="gpt-4o",
        help="The model to use for the user simulator",
    )
    parser.add_argument(
        "--user-model-provider",
        type=str,
        choices=provider_list,
        help="The model provider for the user simulator",
    )
    parser.add_argument(
        "--agent-strategy",
        type=str,
        default="tool-calling",
        choices=["tool-calling", "act", "react", "few-shot"],
    )
    parser.add_argument(
        "--temperature",
        type=float,
        default=0.0,
        help="The sampling temperature for the action model",
    )
    parser.add_argument(
        "--task-split",
        type=str,
        default="test",
        choices=["train", "test", "dev"],
        help="The split of tasks to run (only applies to the retail domain for now",
    )
    parser.add_argument("--start-index", type=int, default=0)
    parser.add_argument("--end-index", type=int, default=-1, help="Run all tasks if -1")
    parser.add_argument(
        "--task-ids",
        type=int,
        nargs="+",
        help="(Optional) run only the tasks with the given IDs",
    )
    parser.add_argument("--log-dir", type=str, default="results")
    parser.add_argument(
        "--max-concurrency",
        type=int,
        default=1,
        help="Number of tasks to run in parallel",
    )
    parser.add_argument("--seed", type=int, default=10)
    parser.add_argument("--shuffle", type=int, default=0)
    parser.add_argument(
        "--user-strategy",
        type=str,
        default="llm",
        choices=[item.value for item in UserStrategy],
    )
    parser.add_argument(
        "--few-shot-displays-path",
        type=str,
        help="Path to a jsonlines file containing few shot displays",
    )
    parser.add_argument(
        "--dataset-path", default="atomwalk12/linalgzero-grpo", type=str, help="Path to the huggingface dataset"
    )
    args = parser.parse_args()
    print(args)
    return RunConfig(
        project_id=args.project_id,
        dataset_path=args.dataset_path,
        project=args.project_name,
        model_provider=args.model_provider,
        model=args.model,
        num_trials=args.num_trials,
        env=args.env,
        agent_strategy=args.agent_strategy,
        temperature=args.temperature,
        task_split=args.task_split,
        start_index=args.start_index,
        end_index=args.end_index,
        task_ids=args.task_ids,
        log_dir=args.log_dir,
        max_concurrency=args.max_concurrency,
        seed=args.seed,
        shuffle=args.shuffle,
        user_strategy=args.user_strategy,
        few_shot_displays_path=args.few_shot_displays_path,
    )


def main():
    config = parse_args()
    run(config)


if __name__ == "__main__":
    main()