-
Notifications
You must be signed in to change notification settings - Fork 78
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
Mesh-Graph Network Example #2185
base: develop
Are you sure you want to change the base?
Conversation
To do:
|
@szaman19 for normalizer you can use batchnorm afaiu |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Great work! Some comments/questions remain on the model and the files need some standard formatting.
@@ -0,0 +1,114 @@ | |||
import lbann |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Missing header (applies to all files)
|
||
|
||
DATA_CONFIG = configparser.ConfigParser() | ||
DATA_CONFIG.read("data_config.ini") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
What if this file does not exist?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ended up including an example file in the example.
name=None): | ||
|
||
super().__init__() | ||
MLP.global_count += 1 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
do we have to give names to MLPs?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I usually add it for compute graph visualization and for better debugging messages when things break
(Layer): Expected shape (Batch, N, self.out_dim) | ||
""" | ||
self.instance += 1 | ||
name = f"{self.name}_instance_{self.instance}" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Unused
class EdgeProcessor(Module): | ||
""" Applies MLP transform on concatenated node and edge features | ||
""" | ||
global_count = 0 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Same
Add synthetic config file
A partially completed data parallel implementation of
Mesh Graph Network
Partially completed as the Normalizer operation in the original is currently unsupported, but potentially could be simply added via an external layer. It is similar to Batch Norm but with some slight changes.
Pinging: @bvanessen @tbennun @ndryden