Can you kindly pinpoint the compression and decompression in code that undergoes in ('aggregated',) tensor tag from previous FL round stored in in db_iterator?

Created by dskhanirfan
@dskhanirfan, I apologize for the miscommunication, in fact 4, 5, and 6 are NOT functions for which customizations will be explored for this year. We will be reflecting this in an updated design doc soon.
@brandon4edwards Following is mentioned in https://zenodo.org/record/6362409#.YoAxVBNBzaa specific instructions will be given to the participants on the parts/functions that they would need to alter the federated algorithm in the following ways: 1. The aggregation function used to fuse the collaborator model updates 2. Which collaborators are chosen to train in each federated round. 3. The training parameters for each collaborator for each federated round. 4. The validation metrics to be computed each round (which can then be used as inputs to the other functions). 5. Compression of model update uploads/downloads. 6. Logic to determine whether to end a round early due to a slow collaborator (this relates to the optional part of Task 1) If compression is not supported this year then What is the 5th point about?
Hello, I will get someone to help you locate the yaml that is used to define the compression pipeline, however I wanted to make sure you understood that utilization of compression is not currently taken into account when computing the simulated run time. That is, we do not support the use of compression in that way for this year. Therefore it will only hurt your performance for this competition. Best, Brandon

compression and decompression in ('aggregated',) tensor page is loading…