Hello i am new to the topic of training a NN on my own so this might be a easy error, but i can’t fix it.
I have the following code
class ClassConditionedUned(nn.Module):
def __init__(self, num_ela=8, class_emb_size=4):
super().__init__()
self.class_emb = nn.Sequential(
nn.Linear(num_ela, 32),
nn.ReLU(),
nn.Linear(32, class_emb_size)
)
self.model = UNet1DModel(
sample_size=512,
in_channels=1+class_emb_size,
out_channels=1,
layers_per_block=1,
block_out_channels = (32, 32, 64),
down_block_types = ("DownBlock1DNoSkip", "DownBlock1D", "AttnDownBlock1D"),
up_block_types = ("AttnUpBlock1D", "UpBlock1D", "UpBlock1DNoSkip"),
)
def forward(self, x, t, ela_vec):
bs, ch, h = x.shape
class_cond = self.class_emb(ela_vec) # Map to embedding dimension
class_cond = class_cond.view(bs, -1, 1).expand(-1, -1, h)
net_input = torch.cat((x, class_cond), 1)
print(net_input.shape)
return self.model(net_input, t).sample
model = ClassConditionedUned()
x = torch.randn(1, 1, 512)
t = torch.randint(0, 1000, (1,))
ela_vec = torch.rand(1, 8) # normalisierte ELA-Vektoren
with torch.no_grad():
out = model(x, t, ela_vec)
and i get the error:
out = model(x, t, ela_vec)
^^^^^^^^^^^^^^^^^^^^
RuntimeError: Given groups=1, weight of size [32, 5, 1], expected input[1, 21, 512] to have 5 channels, but got 21 channels instead
what am i doing wrong?
Thank you in advance!