the sacle of the 3dconv

[loserhou]

in the baseline.py:
self.strides = [int(scale / self.scale_list[0]) for scale in self.scale_list]

for i in range(self.num_scales):
self.spv_enc.append(SPVBlock(
in_channels=self.hiden_size,
out_channels=self.hiden_size,
indice_key='spv_'+ str(i),
scale=self.scale_list[i],
last_scale=self.scale_list[i-1] if i > 0 else 1,
spatial_shape=np.int32(self.spatial_shape // self.strides[i])[::-1].tolist())
)
that means the spatial_shape is 1,2,4,8 not 2,4,8,16?