Use DataLoader in Flux DimensionMismatch

I am trying to train this tiny net in Flux:

	NN = Chain(
		Dense(2 => 16, tanh),
		Dense(16 => 16, tanh),
		Dense(16 => 1)
	) |> f64

I the data comes from a single DataFrame, that I mask, convert the temp to Kelvin and feed this into the DataLoader.
But when I want to train the batches

I get the error: DimensionMismatch: arrays could not be broadcast to a common size; got a dimension with lengths 3 and 64
That the dimensions do not match. Which I do not unterstand.

train_mask = 1: Int.(length(static_df.T1)//2)

x_train = static_df[train_mask, [:Q1, :T1]]
x_train.T1 = x_train.T1 .|> celsius2kelvin
x_train = x_train|> Matrix
x_train  =  [x for x in eachrow(x_train)]

y_train = static_df[train_mask, :T1] .|> celsius2kelvin

loader = Flux.DataLoader((data=x_train, label=y_train), batchsize=64, shuffle=false, partial=false)
η = 0.001
opt = Flux.setup(Adam(η), NN)
epochs = 5
lossₘ(ŷ, y) = sum((ŷ .- y).^2)
for i in 1:epochs
	for mini_batch ∈ loader
		train!(lossₘ, NN, mini_batch, opt)
  1. How can I correctly load to data and
    2.How can I see the progressing loss during the training?