Interval examples throw a dtype warning

[colleenjg]

Many of the Interval object creation examples in the repo specify start and end as lists of ints, e.g.:

from torch_brain.data import Interval

trials=Interval(
    start=[0, 1, 2],
    end=[1, 2, 3],
    go_cue_time=[0.5, 1.5, 2.5],
    drifting_gratings_dir=[0, 45, 90],
)

As a result, they throw the warning:

WARNING:root:start is of type int64 not of type float64.
WARNING:root:end is of type int64 not of type float64.

Two issues:

• This occurs in several locations in the repo:
  • README.md:L78
  • data.py:L62
  • interval.py:L688
  • Several locations in test_data.py, though note L332 which intentionally tests for this warning.
• What the warning refers to may not be obvious to the user if they are creating a complex object.

Suggested fixes:

• Update all examples to use start=[0.0, 1.0, 2.0] and end=[1.0, 2.0, 3.0] (except the warning test line specified above).
• Add a bit of info to the warning (and update what the warning test checks for accordingly):
  logging.warning(f"{name} is of type {value.dtype} not of type float64.") ->
  logging.warning(f"Interval {name} is of type {value.dtype} not of type float64.").

The fixes could be incorporated into the same PR as for #338.

[AlexandreAndr]

Good catch!
We recently supported the use of list/tuples instead of np.ndarray to instanciate the temporaldata objects. During these change I suspect that this kind of changed happened for visual simplification:

start=np.array([0.0, 1.0, 2.0]),

for

start=[0, 1, 2],

Update all examples to use start=[0.0, 1.0, 2.0] and end=[1.0, 2.0, 3.0] (except the warning test line specified above).

Agree!
Minor comment, potentially use start=[0., 1., 2.] instead of start=[0.0, 1.0, 2.0] if it works

Add a bit of info to the warning (and update what the warning test checks for accordingly):
logging.warning(f"{name} is of type {value.dtype} not of type float64.") ->
logging.warning(f"Interval {name} is of type {value.dtype} not of type float64.").

Agree!

[colleenjg]

Great! And sounds good, I'll use start=[0., 1., 2.].