Open braingram opened 2 weeks ago
Refpix when run with numpy 2.0 produces errors:
> offset = output[row] * (nx // 5) # nx // 5 is 640 E OverflowError: Python integer 640 out of bounds for uint8
https://github.com/spacetelescope/jwst/actions/runs/9570099916/job/26384136420?pr=8576#step:10:533
This is due to different type promotion in numpy 2.0.
For numpy < 2
>> np.uint8(6) * 640 3840
For numpy 2:
>> np.uint8(6) * 640 OverflowError: Python integer 640 out of bounds for uint8
The output referred to above appears to be from the IRS2Model with a schema that defines it's datatype as 'uint8': https://github.com/spacetelescope/stdatamodels/blob/18d07d732892ddcdbb8f13098468491edac3d473/src/stdatamodels/jwst/datamodels/schemas/irs2.schema.yaml#L37 This does not match the docstring for the function: https://github.com/spacetelescope/jwst/blob/8c27f16cafc7b3af6e440719c79785e8dc5d40f4/jwst/refpix/irs2_subtract_reference.py#L331-L333
output
Refpix when run with numpy 2.0 produces errors:
https://github.com/spacetelescope/jwst/actions/runs/9570099916/job/26384136420?pr=8576#step:10:533
This is due to different type promotion in numpy 2.0.
For numpy < 2
For numpy 2:
The
output
referred to above appears to be from the IRS2Model with a schema that defines it's datatype as 'uint8': https://github.com/spacetelescope/stdatamodels/blob/18d07d732892ddcdbb8f13098468491edac3d473/src/stdatamodels/jwst/datamodels/schemas/irs2.schema.yaml#L37 This does not match the docstring for the function: https://github.com/spacetelescope/jwst/blob/8c27f16cafc7b3af6e440719c79785e8dc5d40f4/jwst/refpix/irs2_subtract_reference.py#L331-L333