Problem with basic tutorial

[Giulio Tani]

Hi,

I am not able to complete this tutorial : Training a state-of-the-art classifier on the SVHN dataset

I did step 1 and 2 with no problem but in step 3 I have this error:

giulio@giulioBig:~/eblearn/demos/svhn$ train svhn.conf
* Generic trainer
Using fixed seed: 0
Setting conf directory to: ./
_____________________ Configuration _____________________
aft : 1
annea_period : 0
anneal_value : 0.0
answer : class_answer
arch : conv051,addc0,tanh,,l4pool22,snorm5,ms2,merge2,linear6,addc6,tanh,linear7,addc7,tanh
balanced_training : 1
bef : 0
binary_target : 0
c05sh : conv051,addc0,tanh
c17sh : conv171,addc2,tanh
classification : 1
classifier : linear6,addc6,tanh,linear7,addc7,tanh
conv051_kernel : 5x5
conv051_shared : 1
conv051_stride : 1x1
conv051_table : table_3_16_connect_32_fanin_density_0.67_yuv_y8_u0_v0_yuv16_uv0.mat
conv171_kernel : 7x7
conv171_shared : 1
conv171_stride : 1x1
conv171_table : table_16_512_connect_1792_fanin_2_3_4_5_density_0.22_random.mat
current_dir : ./
epoch_mode : 0
epoch_show_modulo : 1000
epoch_size : 10000
eta : .00001
f6 : linear6,addc6,tanh
f7 : linear7,addc7,tanh
features : conv051,addc0,tanh,,l4pool22,snorm5,ms2,merge2
fixed_randomization : 1
gaussian_coeff : 4
gradient_threshold : 0.0
hardest_focus : 1
hessian_period : 20000
ignore_correct : 1
inertia : 0.0
input_thickness : 3
iterations : 1000
keep_outputs : 0
l1 : 1
l1penalty_coeff : .00001
l1penalty_threshold : 0
l4pool22_energy : l1penalty
l4pool22_kernel : 2x2
l4pool22_stride : 2x2
l4pool53_energy : l1penalty
l4pool53_kernel : 5x5
l4pool53_stride : 3x3
learn_mean_norm : 0
learn_norm : 0
linear6_in : 8448
linear6_out : 20
linear7_in : 20
linear7_out : noutputs
merge2_ins : 4x4,4x4
merge2_strides : 1x1,1x1
merge2_type : mflat
merge_padding : 0
min_sample_weight : 0
ms2_pipe0 : conv171,addc2,tanh,,l4pool22,snorm7
ms2_pipe1 : ,l4pool53,snorm5
ndiaghessian : 400
no_testing_test : 0
no_training_test : 1
nonlin : tanh
norm : snorm
norm03 : 
norm05 : 
norm06 : 
norm09 : 
norm15 : snorm5
norm17 : snorm7
norm_div : 0
norm_split : 1
penalty : l1penalty
penalty0 : 
penalty1 : l1penalty
per_class_norm : 1
pooling : l4pool
random_class_order : 1
reg : .00001
reg_l1 : .00001
reg_l2 : .00001
reg_time : 0
root : svhn/
run_type : train
s022_5 : ,l4pool22,snorm5
s122_7 : ,l4pool22,snorm7
s153_5 : ,l4pool53,snorm5
sample_probabilities : 0
save_confusion : 0
save_pickings : 0
save_weights : 1
shared : 1
show_conf : 1
show_hsample : 5
show_train : 0
show_train_errors : 0
show_train_ninternals : 0
show_val_correct : 1
show_val_errors : 1
show_wsample : 18
shrink : lshrink0
shuffle_passes : 1
snorm5_fsum_div : 0
snorm5_fsum_split : 1
snorm5_gaussian_coeff : 4
snorm5_kernel : 5x5
snorm5_learn : 0
snorm7_fsum_div : 0
snorm7_fsum_split : 1
snorm7_gaussian_coeff : 4
snorm7_kernel : 7x7
snorm7_learn : 0
snorm_div : 0
snorm_split : 1
table0 : table_3_16_connect_32_fanin_density_0.67_yuv_y8_u0_v0_yuv16_uv0.mat
table1 : table_16_512_connect_1792_fanin_2_3_4_5_density_0.22_random.mat
target_factor : 1
test_only : 0
train : svhn//svhn_ynuv7_train_data.mat
train_classes : svhn//svhn_ynuv7_train_classes.mat
train_dsname : svhn_ynuv7_train
train_labels : svhn//svhn_ynuv7_train_labels.mat
trainable_module1_energy : l2_energy
trainer : trainable_module1
training_precision : double
val : svhn//svhn_ynuv7_val_data.mat
val_classes : svhn//svhn_ynuv7_val_classes.mat
val_dsname : svhn_ynuv7_val
val_labels : svhn//svhn_ynuv7_val_labels.mat
_________________________________________________________
Training precision: double
Error: failed to open svhn//svhn_ynuv7_val_data.mat, in select_data_type at /home/giulio/eblearn/tools/tools/src/train.cpp:238
Stack:
train(_Z16select_data_typeIdEiRN3ebl13configurationERSs+0x29b)[0x6bb67b]
train(_Z8run_mainiPPc+0x2ac)[0x4a3ebc]
/home/giulio/eblearn/bin/libidxgui.so(_ZN3ebl6idxgui3runEv+0x11)[0x7f335cbb2281]
/usr/lib/x86_64-linux-gnu/libQtCore.so.4(+0x7b32f)[0x7f335c72e32f]
/lib/x86_64-linux-gnu/libpthread.so.0(+0x8182)[0x7f335b5e6182]
/lib/x86_64-linux-gnu/libc.so.6(clone+0x6d)[0x7f335aaf2fbd]
Aborted (core dumped)

Do you now where is the problem??

[EBFun]

Hi Guliano,

Following the tutorial, you are in the step 3 ? (train svhn.conf)

 Open your file "svhn.conf", 

 Check if the "val" variable is assigned the correct path to  "svhn_ynuv7_val_data.mat" that you just dowloaded 

Let me know if you resolve it,

[Giulio Tani]

I found the problem!

The problem was the path of "svhn_ynuv7_val_data.mat".

I moved all the data in "home/giulio/svhn/" and everything worked well :)

The two "//" in the definition of "val" variable confused me.

Thanks EBFun