Logging #143
Logging #143
Conversation
Updated fl.py file to run plot time using tensorboard
Merge to stay up to date
Merging to keep up to date
define self.num_users
src/algos/base_class.py
Outdated
| @@ -339,6 +339,13 @@ def log_metrics(self, stats: Dict[str, Any], iteration: int) -> None: | |||
| imgs=stats["images"], key="sample_images", iteration=iteration | |||
| ) | |||
|
|
|||
| #config["num_users"] | |||
There was a problem hiding this comment.
So is the logic to log a graph at every round or only do it in a posthoc manner (in the plot utils) where we log the graph in the end? I am seeing code in both log_utils and plot_utils so not sure
There was a problem hiding this comment.
the logic is the log the global csv files for each round of each node. Once the final node of a round finishes, the plotter will work to plot the results of that round based on the data produced by the logger.
There was a problem hiding this comment.
But csv files were already getting logged about who is collaborating with whom at every round, why store separate csv files?
There was a problem hiding this comment.
In other words, there isn't any information that needs to be captured in addition to what we were already capturing for generating the three network visualizations (post training), correct?
|
Based on today's discussion, two remaining tasks -
|
Edited Log_utils file to create global graph cvs files and adjacency matricies
Edited plot_utils file to create 3 visualizations based on log_util output files, created a heatmap of node connections, a video of connection progressions throughout rounds and a weighted graphs depicting how often each edge is utilized.