submitit/submitit_train.py [23:57]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
def get_init_file() -> Path:
    # Init file must not exist, but it's parent dir must exist.
    os.makedirs(str(get_shared_folder()), exist_ok=True)
    init_file = get_shared_folder() / f"{uuid.uuid4().hex}_init"
    if init_file.exists():
        os.remove(str(init_file))
    return init_file

def grid_parameters(grid: Dict):
    """
    Yield all combinations of parameters in the grid (as a dict)
    """
    grid_copy = dict(grid)
    # Turn single value in an Iterable
    for k in grid_copy:
        if not isinstance(grid_copy[k], Iterable):
            grid_copy[k] = [grid_copy[k]]
    for p in itertools.product(*grid_copy.values()):
        yield dict(zip(grid.keys(), p))

def grid_search(args):
    cluster_cfg = ClusterConfig(dist_backend="nccl", dist_url="")

    date_curr = date.today().strftime("%m-%d-%Y")
    log_dir = os.path.join(args.output_dir, date_curr)
    
    TrainerConfig = namedtuple("TrainerConfig", sorted(vars(args)))
    train_cfg = TrainerConfig(**vars(args))

    # Create the executor
    print("Create the submitit Executor (can take time on FB cluster)")
    # Note that the folder will depend on the job_id, to easily track experiments
    executor = submitit.AutoExecutor(folder=get_shared_folder() / "%j")
    num_gpus_per_node = 8
    executor.update_parameters(
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



submitit/submitit_train_qa.py [22:56]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
def get_init_file() -> Path:
    # Init file must not exist, but it's parent dir must exist.
    os.makedirs(str(get_shared_folder()), exist_ok=True)
    init_file = get_shared_folder() / f"{uuid.uuid4().hex}_init"
    if init_file.exists():
        os.remove(str(init_file))
    return init_file

def grid_parameters(grid: Dict):
    """
    Yield all combinations of parameters in the grid (as a dict)
    """
    grid_copy = dict(grid)
    # Turn single value in an Iterable
    for k in grid_copy:
        if not isinstance(grid_copy[k], Iterable):
            grid_copy[k] = [grid_copy[k]]
    for p in itertools.product(*grid_copy.values()):
        yield dict(zip(grid.keys(), p))

def grid_search(args):
    cluster_cfg = ClusterConfig(dist_backend="nccl", dist_url="")

    date_curr = date.today().strftime("%m-%d-%Y")
    log_dir = os.path.join(args.output_dir, date_curr)
    
    TrainerConfig = namedtuple("TrainerConfig", sorted(vars(args)))
    train_cfg = TrainerConfig(**vars(args))

    # Create the executor
    print("Create the submitit Executor (can take time on FB cluster)")
    # Note that the folder will depend on the job_id, to easily track experiments
    executor = submitit.AutoExecutor(folder=get_shared_folder() / "%j")
    num_gpus_per_node = 8
    executor.update_parameters(
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -