search

NVIDIA / DIGITS

Deep Learning GPU Training System
https://developer.nvidia.com/digits
BSD 3-Clause "New" or "Revised" License
4.12k stars 1.38k forks source link

ValueError: could not broadcast input array from shape (51,5) into shape (50,5) #1682

Open caleale90 opened 7 years ago

caleale90 commented 7 years ago

Hi, I am facing the same error of the #1235 issue. I followed the @lukeyeager instruction and changed the variable MAX_BOXES = 150 in both files clustering.py and mean_ap.py. Then I recompiled the entire caffe library, but I obtain the same error. This is the last part of my caffe_output.log file:

I0616 09:35:50.296716 5248 sgd_solver.cpp:106] Iteration 20000, lr = 0.0001 I0616 09:40:43.958341 5248 solver.cpp:362] Iteration 20759, Testing net (#0) I0616 09:40:43.958425 5248 net.cpp:723] Ignoring source layer train-data I0616 09:40:43.958427 5248 net.cpp:723] Ignoring source layer train-label I0616 09:40:43.958429 5248 net.cpp:723] Ignoring source layer train-transform Traceback (most recent call last): File "/usr/lib/python2.7/dist-packages/caffe/layers/detectnet/clustering.py", line 133, in forward bbox = cluster(self, data0, bottom[1].data) File "/usr/lib/python2.7/dist-packages/caffe/layers/detectnet/clustering.py", line 232, in cluster boxes[i, 0:r, 0:c] = boxes_cur_image ValueError: could not broadcast input array from shape (51,5) into shape (50,5)

Can anyone help me to solve this problem?

ontheway16 commented 7 years ago

Yes. :)

Actually I have experienced this problem for a long time and my solution was defining necessary paths like DIGITS_ROOT and CAFFE_ROOT. Cannot remember now if there were other paths to define, but try these two first.

caleale90 commented 7 years ago

How did you set these two variables? Which value did you put inside?

ontheway16 commented 7 years ago

see here: #1440