Yyk040316/long-context-icl

1import subprocess2import argparse3# 定义数据集、任务和 n_shots4 5 6parser = argparse.ArgumentParser(description="Run experiments with specified models and tensor parallel size.")7parser.add_argument('--models-path', nargs='+', required=True, help="Path(s) to the models.")8parser.add_argument('--tensor-parallel-size', nargs = '+',type=int, required=True, help="Tensor parallel size.")9args = parser.parse_args()10 11 12dataset_task_map = {13    "Multilingual_Bemba": "multilingual",14    "Multilingual_French": "multilingual",15    "Multilingual_German": "multilingual",16    "Multilingual_Kurdish": "multilingual",17    "News": "summarization",18    "Bill": "summarization",19    "Dialogue": "summarization",20    "Intent": "classification",21    "Topic": "classification",22    "Sentiment": "classification",23    "Marker": "classification",24    "Commonsense": "qa",25    "Science": "qa",26    "Medical": "qa",27    "Retrieval": "qa",28    "Law": "qa"29    30}31 32dataset_shots_map = {33    "Multilingual_Bemba": [1, 5, 25, 50, 100, 200, 500, 800, 1000],34    "Multilingual_French": [1, 5, 25, 50, 100, 200, 500, 800, 1000],35    "Multilingual_German": [1, 5, 25, 50, 100, 200, 500, 800, 1000],36    "Multilingual_Kurdish": [1, 5, 25, 50, 100, 200, 500, 800, 1000],37    "News": [1, 5, 10, 25, 50, 75, 100, 150, 200],38    "Bill": [1, 5, 10, 25, 30, 35, 40, 45, 50],39    "Dialogue": [1, 5, 10, 25, 50, 100, 200, 300, 400, 500],40    "Intent": [1, 5, 10, 25, 50, 100, 200, 500, 800,1000,2000],41    "Topic": [1, 5, 10, 25, 50, 100, 200, 500, 800],42    "Sentiment": [1, 5, 10, 25, 50, 100, 200, 300, 400, 500],43    "Marker": [1, 5, 10, 25, 50, 100, 200, 500, 800, 1000],44    "Commonsense": [1, 5, 10, 25, 50, 100, 200, 500, 800, 1000],45    "Science": [1, 5, 10, 25, 50, 75, 100, 150, 200],46    "Medical": [1, 5,10, 25, 50, 100, 200, 300, 400, 500],47    "Retrieval": [1, 5, 10, 25, 50, 100, 150, 200],48    "Law": [1, 5, 10, 25, 50, 100,200,250]49 50    51}52 53 54# 定义公共参数55models_path = args.models_path56output_dir = "/path/to/output"57random_seed = 4358n_runs = 559tensor_parallel_size_list = args.tensor_parallel_size60 61 62 63 64for dataset, task in dataset_task_map.items():65    num = 066    for model_path in models_path:67        n_shots = [str(i) for i in dataset_shots_map[dataset]]68        output_dir = f"./{dataset}"69        tensor_parallel_size = tensor_parallel_size_list[num]70        num += 171        72        command = [73            "python3", "./Integrate_Code/main.py",74            "--datasets", dataset,75            "--models-path", model_path,76            "--output-dir", output_dir,77            "--random-seed", str(random_seed),78            "--n-runs", str(n_runs),79            "--n-shots"80        ] + n_shots + [  # 将 n_shots 列表展开为独立的参数81            "--gpu-num", str(tensor_parallel_size),82            "--task", task83        ]84        print(f"Running command: {' '.join(command)}")85        try:86            subprocess.run(command, check=True)87        except subprocess.CalledProcessError as e:88            print(f"命令执行失败: {e}. 跳过并继续下一个命令。")89        90 91