About" invalid output_size "torch.Size([4, 45, 80])" (dim 1 must be between 41 and 45)"

[lianyingjieL]

When I run the original code, the following problems occur：
NVIDIA RTX A2000 x 2，Python3.7, CUDA 11.1, cuDNN 8.0.5, Pytorch 1.8.1, Horovod 0.22.1, Ubuntu 18.04, VNC, NVCC11.1, OpenMPI 4.0.0。

(myconda) root@b3193a16b028:/mnt/lyj/TASED_NET# python run_train.py
loading weight file
loaded
32
torch.Size([4, 3, 32, 360, 640])
torch.Size([4, 3, 32, 360, 640])
Traceback (most recent call last):
File "run_train.py", line 119, in
main()
File "run_train.py", line 92, in main
output = model(clip.cuda())
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 167, in forward
outputs = self.parallel_apply(replicas, inputs, kwargs)
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 177, in parallel_apply
return parallel_apply(replicas, inputs, kwargs, self.device_ids[:len(replicas)])
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/nn/parallel/parallel_apply.py", line 86, in parallel_apply
output.reraise()
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/_utils.py", line 429, in reraise
raise self.exc_type(msg)
ValueError: Caught ValueError in replica 0 on device 0.
Original Traceback (most recent call last):
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/nn/parallel/parallel_apply.py", line 61, in _worker
output = module(*input, **kwargs)
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/mnt/lyj/TASED_NET/model.py", line 98, in forward
z = self.unpool2(z, i1, y2.size())
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/nn/modules/pooling.py", line 462, in forward
self.padding, output_size)
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/nn/functional.py", line 843, in max_unpool3d
output_size = _unpool_output_size(input, kernel_size, _stride, padding, output_size)
File "/root/miniconda3/envs/myconda/lib/python3.7/site-packages/torch/nn/functional.py", line 754, in _unpool_output_size
output_size, d, min_size, max_size
ValueError: invalid output_size "torch.Size([4, 45, 80])" (dim 1 must be between 41 and 45)

[kylemin]

Since TASED-Net has a series of max-pooling layers, it assumes that the spatial dimension of input (width and height) is a multiple of 16. I think you can resize your input to 640x320. As described in our paper, we actually resized every frame to 384x224.