search

lutzroeder / netron

Visualizer for neural network, deep learning and machine learning models
https://netron.app
MIT License
27.84k stars 2.76k forks source link

Isolated blocks in network visualisation #1235

Closed skrishnanv closed 7 months ago

skrishnanv commented 7 months ago

hello, thanks for the Netron. I have a doubt regarding the visualisation. I do not think it is an "issue". However, there seems to be no other way to discuss this than raise a issue. Attached is an image of a simple cnn with standard blocks of conv, downsample, non-lin, a kind of residual connections etc. My doubt is: at the top why are the additional add_ blocks and two modules rendered? i have tried to see their input and output numbers but unable to map them at a quick glance. i have rendered them in tensorboard and there i do not see these isolated blocks. However, i prefer to use netron. so, any explanation to resolve this confusion will be deeply helpful. thanks

model_scripted

lutzroeder commented 7 months ago

@skrishnanv can you share a sample file to reproduce the issue and did you investigate why this is happening.

skrishnanv commented 7 months ago

Hi @lutzroeder , i will get this over the next weekend and post here. thanks!