Possibility to support python-fire?

[TheShadow29]

I usually use python fire (https://github.com/google/python-fire), it creates the parser arguments by default looking at the function to be called. Is there any possibility of integrating this into the existing framework?

[williamFalcon]

what do you mean usually? looks cool, can you share an example?

[TheShadow29]

Here is an example

"""
main.py
"""
import fire
from utils import get_data

def main(exp_name, batch_size, num_workers):
    print(f'Starting Experiment {exp_name}')
    data = get_data(batch_size, num_workers)

if __name__ == '__main__':
    fire.Fire(main)

Now, in my terminal, I can directly do this

python main.py "exp1" 32 8

This would directly pass the arguments to the main function without explicitly importing argparse and is more succint. To be specific, it is like a wrapper around the argparse and calls it internally.

In general, fire supports many other stuff like here: https://github.com/google/python-fire/blob/master/docs/guide.md. Though not all of it would be relevant to test-tube.

[williamFalcon]

Interesting.... could you propose an example using test tube (ie: a dream use case)? I see this as an addition and not necessarily a replacement of argparse because most academic researchers use argparse right now so it's easy to integrate with their code (ie... not many changes)

[TheShadow29]

Yeah, I meant this to be an addition. Here is a small example I am cooking up from your example. https://github.com/williamFalcon/test-tube/blob/master/examples/pytorch_hpc_example.py

The given code:

# set up our argparser and make the y_val tunable
parser = HyperOptArgumentParser(strategy='random_search')
parser.add_argument('--test_tube_exp_name', default='my_test')
parser.add_argument('--log_path', default='/some/path/to/log')
parser.opt_list('--y_val', default=12, options=[1, 2, 3, 4, 5, 6], tunable=True)
parser.opt_list('--x_val', default=12, options=[20, 12, 30, 45], tunable=True)
hyperparams = parser.parse_args()

can be rewritten as:

"""eg.py"""
def main(*args, **kwargs):
     # cfg.json has a dictionary like {"strategy": "random_search"} and other default values
    cfg = json.load(open('./cfg.json'))
    cfg.update(kwargs)
    # Use cfg to update hyperopt
    pass
if __name__ = '__main__':
      fire.Fire(main)

Now from my terminal I can simply call, python eg.py --y_val=5 and done. Makes the code way simpler I think.

[dreamgonfly]

There are other cool arguments parser libraries like http://docopt.org/ as well. We could generalize this issue by supporting json format parsed arguments, instead of relying on a specific format.