Can't test the trained encoder

[mees]

Hi,
I trained the single class encoder with ./demo_pretrain_singleclass.sh. Now I wanted to evaluate the trained models. So in eval_quant_test.lua I just changed the name of the loaded file (cnn_vol.t7) to the last trained model:
base_loader = torch.load(opt.checkpoint_dir .. 'arch_rotatorRNN_singleclass_nv24_adam2_bs8_nz512_wd0.001_lbg10_ks16/net-epoch-20.t7')
encoder = base_loader.encoder
base_voxel_dec = base_loader.voxel_dec

When I run the testcript eval_models.sh I get following error:
/home/meeso/torch/install/bin/luajit: scripts/eval_quant_test.lua:90: attempt to index global 'base_voxel_dec' (a nil value)
stack traceback:
scripts/eval_quant_test.lua:90: in main chunk
[C]: in function 'dofile'
...eeso/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
[C]: at 0x00406670

Any idea how I fix this?

[xcyan]

You need to train a voxel decoder for test.

…

On Fri, Jun 2, 2017 at 8:29 AM Oier Mees ***@***.***> wrote: Hi, I trained the single class encoder with ./demo_pretrain_singleclass.sh. Now I wanted to evaluate the trained models. So in eval_quant_test.lua I just changed the name of the loaded file (cnn_vol.t7) to: base_loader = torch.load(opt.checkpoint_dir .. 'arch_rotatorRNN_singleclass_nv24_adam2_bs8_nz512_wd0.001_lbg10_ks16/net-epoch-20.t7') encoder = base_loader.encoder base_voxel_dec = base_loader.voxel_dec When I run the testcript eval_models.sh I get following error: /home/meeso/torch/install/bin/luajit: scripts/eval_quant_test.lua:90: attempt to index global 'base_voxel_dec' (a nil value) stack traceback: scripts/eval_quant_test.lua:90: in main chunk [C]: in function 'dofile' ...eeso/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk [C]: at 0x00406670 — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#3>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AIWuK0QY947UXFv801ohduGmsS8DC1qPks5sACo3gaJpZM4Nubz7> .

[mees]

Ah okey you are right. But it seems that in the code the encoder doesn't get loaded:

nips16_PTN/scripts/train_PTN.lua

Line 108 in 789031e

opt.basemodel_name = string.format('%s_%s_nv%d_adam%d_bs%d_nz%d_wd%g_lbg%g_ks%d',

Not sure if its the commented out lines later, but I think a torch.load() is missing somewhere in the code right? Or how can I tell him to use my trained encoder for the training?

Also, is there some method for storing a log with the train and test loss? At least this way one could see if the trained encoder is performing, or how do you monitor it?

[xcyan]

You are right. This should be fixed. Add back the two lines: loader = torch.load(opt.checkpoint_dir .. opt.basemodel_name .. string.format('/net-epoch-%d.t7', opt.basemodel_epoch)) encoder = loader.encoder

…

On Fri, Jun 2, 2017 at 9:28 AM Oier Mees ***@***.***> wrote: Ah okey you are right. But it seems that in the code the encoder doesn't get loaded: https://github.com/xcyan/nips16_PTN/blob/789031e2c6d63648a1debc2134dd7fd525bf735f/scripts/train_PTN.lua#L108 Not sure if its the commented out lines later, but I think a torch.load() is missing somewhere in the code right? Or how can I tell him to use my trained encoder for the training? Also, is there some method for storing a log with the train and test loss? At least this way one could see if the trained encoder is performing, or how do you monitor it? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#3 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AIWuK5URc2FSlJmH3Ke4-79UOL_LLXCyks5sADg0gaJpZM4Nubz7> .

[mees]

Thanks, I also had to change the name 'rotatorRNN1_64' to 'arch_rotatorRNN'

[xcyan]

Fixed the bug reported by @mees
8cf26ed