When using the three loggers in LFW notebook, Intercluster Metrics is the one that takes the most of the time.
Also, seeing wandb logs, we can see that GPU utilization is almost zero when computing these values. In #35 we have to see which parts are the most critical to optimize, but seems like this will be one.
When using the three loggers in LFW notebook, Intercluster Metrics is the one that takes the most of the time.
Also, seeing wandb logs, we can see that GPU utilization is almost zero when computing these values. In #35 we have to see which parts are the most critical to optimize, but seems like this will be one.