mask is not applied to output NetCDF file if remove_mean = false

[Alexander-Barth]

Describe the bug
mask is not applied to output NetCDF file if remove_mean = false

To Reproduce

using Test
using DINCAE
using Base.Iterators
using Random
using NCDatasets
using CUDA

const F = Float32
Atype =
    if CUDA.functional()
        CuArray{F}
    else
        Array{F}
    end

Random.seed!(123)

filename = "/home/abarth/Downloads/ds_test.nc"

data = [
    (filename = filename,
     varname = "CHL",
#     varname = "DINO",
     obs_err_std = 1,
     jitter_std = 0.05,
     isoutput = true,
     ),
    (filename = filename,
#     varname = "CHL",
     varname = "DINO",
     obs_err_std = 1,
     jitter_std = 0.05,
     isoutput = true,
     )

]
data_test = data;
data_all = [data,data_test]

epochs = 3
batch_size = 5
save_each = 10
skipconnections = [1,2]
enc_nfilter_internal = round.(Int,32 * 2 .^ (0:3))
clip_grad = 5.0
regularization_L2_beta = 0
save_epochs = [epochs]
is3D = false
ntime_win = 3

(upsampling_method,is3D,truth_uncertain,loss_weights_refine) = (:nearest, false,false, (1.,))


fnames_rec = ["ehsan_out.nc"]
paramfile = "ehsan_params.nc"

losses = DINCAE.reconstruct(
    Atype,data_all,fnames_rec;
    epochs = epochs,
    batch_size = batch_size,
    truth_uncertain = truth_uncertain,
    enc_nfilter_internal = enc_nfilter_internal,
    clip_grad = clip_grad,
    save_epochs = save_epochs,
    is3D = is3D,
    upsampling_method = upsampling_method,
    ntime_win = ntime_win,
    loss_weights_refine = loss_weights_refine,
    paramfile = paramfile,
    remove_mean = false,
)


ds = NCDataset(fnames_rec[1])
lon = ds["lon"][:]
lat = ds["lat"][:]
data = nomissing(ds["DINO"][:,:,1],NaN) # change missing to NaN

# plot the first time instant
using PyPlot # you can use python directly of course
pcolor(lon,lat,data')
savefig("/tmp/ehsan.png")

Expected behavior

Spain (and others) should be masked.

Environment

• DINCAE version

bdc9c94

Full output

(no error)

CC @EhsanMehdipour

[Alexander-Barth]

@EhsanMehdipour does the current fix work for you? (in any case, the output of the values on sea are still ok).

[EhsanMehdipour]

updating the repository and packages fixed the issue for the test dataset.

[EhsanMehdipour]

to ensure that the CUDA toolkit is working on the HPC system, I had to set the CUDA runtime version by CUDA.set_runtime_version!(v"12.1.1")