-
Notifications
You must be signed in to change notification settings - Fork 1.4k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
More graph visualization #1424
Comments
This issue has been automatically marked as stale because it has not had recent activity. It will be closed after 30 days if no further activity occurs. Thank you for your contributions. |
This issue is closed as announced. Feel free to re-open it if needed. |
This issue has been automatically marked as stale because it has not had recent activity. It will be closed after 30 days if no further activity occurs. Thank you for your contributions. |
This may be useful also for other purposes like ONNX export. Let's discuss the unified interface of such annotation feature. |
Chainer currently has a feature of drawing the computational graph using graphviz, but the resulting graph is hard to dig into when the graph is very large and complicated. Relevant information like weight and activation statistics is also not shown in the graph. It is great if we have an interactive view of the computational graph with relevant information. This topic is influenced by TensorBoard, the official visualization tool for TensorFlow.
Supporting such a rich visualization on Chainer is not straight-forward, because Chainer does not build any graphs before training. The following list is a set of features possibly required for the visualization:
with
statement or decorator) to specify a meaningful subgraph. We also have to provide a natural and standard way to reflect the link/chain structure into the computational graph by annotations.See also the discussion in the forum: https://groups.google.com/forum/#!topic/chainer/GqXjtHCgQgE
The text was updated successfully, but these errors were encountered: