The EAGLE
directory contains codes for reproducing our experiments.
Directory newrules
contains code for the new rules for TorchRec and other new applications.
To create the docker container, first enter the EAGLE
directory, then run bash docker_command
.
To generate and save input datasets, under the EAGLE
directory, run python -m newrules.gen_torchrec_model_and_dataset
. Input will be saved under EAGLE/data/dataset
and models will be saved under EAGLE/data/models
.
To execute the new rules regarding distributed versus non-distributed training and inference (e.g., rule 17), run bash execute_testing_new_rules.sh
. The outputs will be saved under EAGLE/data/outputs
.
After execution, use python analyze_results_distributed.py
to analyze the results for the new rules regarding distributed versus non-distributed training and inference (e.g., rule 17).