Error data shape mismatched for unaligned data set

[nhnkkhang]

Hello team,
I tried to create my own "unaligned" data set with branch data size (N, 1000), trunk data size (N,50), and output data size (N,50). And build the model as follows:
data = dde.data.triple.Triple(X_train=x_train, y_train=y_train, X_test=x_test, y_test=y_test)
npoints =50
m = 1000
net = dde.nn.DeepONet(
[m, 100,100,100,100],
[npoints, 100,100,100,100], "relu",
"Glorot normal"
)
model = dde.Model(data, net)
model.compile(
optimizer="adam",
lr=1e-3,
loss='mse',
metrics=["mse"]
)
losshistory, train_state = model.train(epochs=500000, batch_size=batch_size, model_save_path="./mdls/PKE_rDONtest_"+str(points))

But I keep getting this error message: ValueError: Cannot feed value of shape (32, 50) for Tensor Placeholder_74:0, which has shape (None, 1). Why does my output in the model expect a size of (None,1) but not (None,50)?
Could you please help me resolve this?
Thank you

[nhnkkhang]

@lululxvi Could you please help me with this? I tried to build my own DeepONet with pytorch to solve this problem but the performance is pretty low.

[voloddia]

For unaligned data (i.e. not CartesianProduct), DeepONet expects each sample to have one coordinate in space/time in trunk_net input. So you need to rearrange your input accordingly.