[BUG]AttributeError: 'Sequential' object has no attribute '_layers'

[JuggernautPanda]

import visualkeras model.build() visualkeras.layered_view(model, legend=True, to_file='output.png').show() # write and show visualkeras.graph_view(model,show_neurons=True, to_file='graph.png').show()

AttributeError Traceback (most recent call last)

in () 2 model.build() 3 visualkeras.layered_view(model, legend=True, to_file='output.png').show() # write and show ----> 4 visualkeras.graph_view(model,show_neurons=True, to_file='graph.png').show() 1 frames /usr/local/lib/python3.7/dist-packages/visualkeras/layer_utils.py in model_to_adj_matrix(model) 37 if not model.built: 38 model.build() ---> 39 layers = model._layers 40 41 adj_matrix = np.zeros((len(layers), len(layers))) AttributeError: 'Sequential' object has no attribute '_layers'

[JuggernautPanda]

I tried the fix by changing the _layers to layers.

I get a new issue:

/usr/local/lib/python3.7/dist-packages/visualkeras/layer_utils.py in model_to_adj_matrix(model) 55 src = id_to_num_mapping[inbound_layer_id] 56 tgt = id_to_num_mapping[layer_id] ---> 57 adj_matrix[src, tgt] += 1 58 59 return id_to_num_mapping, adj_matrix

IndexError: index 12 is out of bounds for axis 1 with size 12

[github-actions[bot]]

Stale issue message

[paulgavrikov]

What tensorflow version are you using? Sounds like an older one. _layers contains layers that tensorflow implicitly adds, so I'd assume that when you forgot to change it somewhere.

[alessiogmonti]

What tensorflow version are you using? Sounds like an older one. _layers contains layers that tensorflow implicitly adds, so I'd assume that when you forgot to change it somewhere.

Hey- I'm having the same 'Sequential' object has no attribute '_layers' issue. I'm using tensorflow 2.7. Any idea what it might be?

Changing _layers to layers returns a similar error:

/usr/local/lib/python3.7/dist-packages/visualkeras/layer_utils.py in model_to_adj_matrix(model)
     55             src = id_to_num_mapping[inbound_layer_id]
     56             tgt = id_to_num_mapping[layer_id]
---> 57             adj_matrix[src, tgt] += 1
     58 
     59     return id_to_num_mapping, adj_matrix

IndexError: index 4 is out of bounds for axis 1 with size 4

This is a great library- looks amazing, but it would be a shame if the graph view doesn't work- I'm not a very experienced programmer but if there is any way I can help please let me know.

[paulgavrikov]

@alessiogmonti I will take a look at this, but due to the upcoming CVPR deadline it will take a while. It looks like the _layers attribute has been abandoned and seems to have been replaced by _self_tracked_trackables in recent tensorflow versions. Could you try changing it? Also, does the bug only affect graph_view or also layered_view?

[alessiogmonti]

@paulgavrikov replacing _layers with _self_tracked_trackables works great- thank you! It runs just fine now. This bug only affects graph_view.

[paulgavrikov]

@alessiogmonti awesome. Would you like to try and create a PR that fixes this? You could check if the passed model has the _layers arg and fallback to the other solution if not. Ideally, we’d also integrate a test into Travis CI

[alessiogmonti]

@paulgavrikov Absolutely- I'll do this today.

[jainakhil88]

@alessiogmonti awesome. Would you like to try and create a PR that fixes this? You could check if the passed model has the _layers arg and fallback to the other solution if not. Ideally, we’d also integrate a test into Travis CI

Any updates on this, is it fixed? I am getting below error

[alessiogmonti]

Hey, for now you can use this. It will try to use _layers and fall back to _self_tracked_trackables when necessary: !pip install git+https://github.com/alessiogmonti/visualkeras

[paulgavrikov]

The PR has been merged into the master branch. @jainakhil88 Your issue is related to something else. It looks like you are not passing a valid Keras model to visualkeras. Are you sure MaskRCNN is not wrapping the actual model?

[github-actions[bot]]

Stale issue message