htoyryla
commented
6 years ago @htoyryla I read you supplied an empty array for the style layers. Could you please tell me how you did that?
No idea, after all this time. I guess I modified the code so that if no style layers are given, it uses an empty array. But reading what I wrote back then, the training failed (loss not decreasing). I think there are several viable solutions for superresolution.
I think I have used these (in no particular order):
https://github.com/alexjc/neural-enhance https://github.com/jacobgil/pytorch-zssr https://github.com/pytorch/examples/tree/master/super_resolution
suke27
I am attempting to train a super-resolution model, but I am getting errors below (using a 4x dataset prepped with make_sr_dataset.py).
@htoyryla I read you supplied an empty array for the style layers. Could you please tell me how you did that?
The training command I am running is:
th train.lua -h5_file superRes_trainSet_284_4x_A.h5 -task upsample -checkpoint_name upsample_checkpoint_A_5k -gpu 0 -num_iterations 5000 -batch_size 1 -arch c9s1-64,R64,R64,R64,R64,u64,c9s1-3What would a valid architecture be for 4x?
Might anyone be willing to share a super-res model just so I can see if it is even worth further investigating training my own model? (2x or 4x would be perfect).
/home/prime/torch/install/bin/luajit: /home/prime/torch/install/share/lua/5.1/nn/Container.lua:67:In 17 module of nn.Sequential:/home/prime/torch/install/share/lua/5.1/nn/THNN.lua:109: inconsistent tensor size at /home/prime/torch/extra/nn/lib/THNN/generic/MSECriterion.c:17stack traceback:[C]: in function 'v'/home/prime/torch/install/share/lua/5.1/nn/THNN.lua:109: in function 'MSECriterion_updateOutput'/home/prime/torch/install/share/lua/5.1/nn/MSECriterion.lua:14: in function 'forward'./fast_neural_style/ContentLoss.lua:43: in function <./fast_neural_style/ContentLoss.lua:39>[C]: in function 'xpcall'/home/prime/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'/home/prime/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward' ./fast_neural_style/PerceptualCriterion.lua:134: in function 'forward'train.lua:197: in function 'opfunc'/home/prime/torch/install/share/lua/5.1/optim/adam.lua:37: in function 'adam'train.lua:239: in function 'main'train.lua:327: in main chunk[C]: in function 'dofile'...rime/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk [C]: at 0x00405d50WARNING: If you see a stack trace below, it doesn't point to the place where this error occured. Please use only the one above.stack traceback:[C]: in function 'error'/home/prime/torch/install/share/lua/5.1/nn/Container.lua:67: in function 'rethrowErrors'/home/prime/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'./fast_neural_style/PerceptualCriterion.lua:134: in function 'forward'train.lua:197: in function 'opfunc'/home/prime/torch/install/share/lua/5.1/optim/adam.lua:37: in function 'adam'train.lua:239: in function 'main'train.lua:327: in main chunk[C]: in function 'dofile'...rime/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk[C]: at 0x00405d50