1

guys, I download resnet-152 from this url: http://data.mxnet.io/models/imagenet-11k/resnet-152/

Then, I try to plot this resnet with 152 layers in this way:

import mxnet as mx 

sym, arg_params, aux_params = mx.model.load_checkpoint('resnet-152', 0)
pltn = mx.viz.plot_network(sym, shape = {'data': (1, 3, 224, 224)}, \
node_attrs = {'shape': 'oval', 'fixedsize': 'false'})
pltn.view('resnet-152')

However, the output pdf file is empty. I can plot LeNet, AlexNet and VGG, but what's wrong with the resnet? Thank you very much.

tsuifrank
  • 11
  • 2
  • Your code works fine for me. Generates a 150k PDF, so it's not huge. Are you sure the PDF is actually blank? The upper-left part of the image (likely default view) is all white, because the giant network gets rendered from bottom-left to top-right of the page. – Leopd Jan 11 '17 at 01:25
  • Thank you very much. It has some relationship with "save_format". If I set save_format = 'eps', I can see the plot. However, the plot just shows part of the network, maybe it's because the network is too large. Can you see the whole neural network? Thanks. – tsuifrank Jan 13 '17 at 08:27

1 Answers1

0

I can confirm that the code as written works with MXNet 1.1.0 and outputs a PDF that contains a graph for the whole neural network.

Sina Afrooze
  • 960
  • 6
  • 11