minor

        env_name = value.get('env_name')

        custom_env_config = value.get('env_config', {})

        train_size,test_size = (value.get('train_size', 100), value.get('test_size', 100))

        env_size = train_size + test_size

        env_config = REGISTERED_ENV[env_name]["config_cls"](**custom_env_config)

        seeds_for_env = None

        seeds_for_env_train = None

        seeds_for_env_test = None

        if hasattr(env_config, 'generate_seeds'):

            seeds_for_env = env_config.generate_seeds(env_size)

            print(f"Using {len(seeds_for_env)} seeds generated by {env_name} config's generate_seeds method")

            seeds_for_env_train = env_config.generate_seeds(train_size)

            seeds_for_env_test = env_config.generate_seeds(test_size)

            print(f"Using {len(seeds_for_env_train)} trian seeds generated by {env_name} config's generate_seeds method")

            print(f"Using {len(seeds_for_env_test)} test seeds generated by {env_name} config's generate_seeds method")

        else:

            seeds_for_env = np.random.randint(0, 2**31 - 1, size=env_size).tolist()

        for seed in seeds_for_env[:train_size]:

            seeds_for_env_train = np.random.randint(0, 2**31 - 1, size=train_size).tolist()

            seeds_for_env_test = np.random.randint(0, 2**31 - 1, size=test_size).tolist()

        for seed in seeds_for_env_train:

            env_settings = {

                'env_name': env_name,

                'env_config': custom_env_config,

                "extra_info": {"split": "train", **env_settings}

            }

            train_instances.append(instance)

        for seed in seeds_for_env[train_size:]:

        for seed in seeds_for_env_test:

            env_settings = {

                'env_name': env_name,

                'env_config': custom_env_config,