Allow dictionary for overrides in `launch`

[jgbos]

Currently launch requires overrides to be a list of strings of type param=value.

This PR allows the user to provide a dictionary as an override: {param: val, ...}. Where param is a string and val can be one of int, float, str, and bool along with with two special types:

• hydra_list: A list type supported by Hydra, e.g., hydra_list([1,2,3]) -> "[1,2,3]"
• multirun: A list that will be converted into a Hydra multirun value, e.g., multirun([1,2,3]) -> "1,2,3"

Override Types

override_primitive_examples = {
    "str": "\(1,2\)",   # --> "\(1,2\)" (no change)
    "NULL": None,   # --> "null"
    "int": 10,  #  --> "10"
    "float" 3.14,  # --> "3.14"
    "bool": True,    # --> "True"
    "list": hydra_list([1, 2, 3]),  # --> "[1, 2, 3]"
    "multirun_list": multirun([1, 2, 3])  # --> "1,2,3"
}

Example of multirun

The use of the multirun type is what really makes this PR special. With the current overrides structure a user must create the string that contains the list of values to multirun overview.

import random
from hydra_zen import launch, instantiate, make_config

values_for_experiment = [random.uniform(0, 1) for i in range(10)]
ov = []

jobs = launch(
    make_config(),
    instantiate,
    overrides=[
         "+param=" + ",".join([str(i) for i in values_for_experiment])
    ],
    multirun=True
)

This PR allows the user to provide the multirun value without having to create the string:

import random
from hydra_zen import launch, instantiate, make_config, multirun

values_for_experiment = [random.uniform(0, 1) for i in range(10)]

jobs = launch(
    make_config(),
    instantiate,
    overrides={
         "+param": multirun(values_for_experiment)
    },
    multirun=True
)

TODO:

• [ ] Update this PR description
• [ ] Add null support in supported types
• [ ] Test configs are the same for list of strings and dictionary
• [ ] Update docstrings
• [ ] Update docs