BUTTER-E - Energy Consumption Data for the BUTTER Empirical Deep Learning Dataset

Description

The BUTTER-E - Energy Consumption Data for the BUTTER Empirical Deep Learning Dataset adds node-level energy consumption data from watt-meters to the primary sweep of the BUTTER - Empirical Deep Learning Dataset. This dataset contains energy consumption and performance data from 63,527 individual experimental runs spanning 30,582 distinct configurations: 13 datasets, 20 sizes (number of trainable parameters), 8 network "shapes", and 14 depths on both CPU and GPU hardware collected using node-level watt-meters. This dataset reveals the complex relationship between dataset size, network structure, and energy use, and highlights the impact of cache effects. BUTTER-E is intended to be joined with the BUTTER dataset (see "BUTTER - Empirical Deep Learning Dataset on OEDI" resource below) which characterizes the performance of 483k distinct fully connected neural networks but does not include energy measurements.

Resources

Name Format Description Link
45 Training losses related to the BUTTER-E dataset, re-summarized from the BUTTER dataset. https://data.openei.org/files/5991/summary_by_epoch.tar
28 README document describing the columns, schema, size, and format of the data contained in this submission. https://github.com/NREL/BUTTER-E-Empirical-analysis-of-energy-trends-in-neural-networks-supplementary-code/blob/main/Readme%20for%20Data.md
0 Paper detailing the BUTTER-E project and dataset. https://arxiv.org/html/2403.08151v1#S3
28 README document describing the columns, schema, size, and format of the data contained in this submission. https://data.openei.org/files/5991/Readme%20for%20Data.md
21 Link to the OEDI submission for the BUTTER dataset which includes a link to the original BUTTER data on AWS, data descriptions, and a tutorial Jupyter notebook for using the data. https://data.openei.org/submissions/5708
57 1-minute raw time series power data corresponding to the runs in the "BUTTER-E Metadata" resource. https://data.openei.org/files/5991/butter_e_energy.zip
8 Characteristics of each compute node used to generate the BUTTER-E data set. https://data.openei.org/files/5991/node_sinfo.csv
8 Power consumption quantiles for each node used to generate the BUTTER-E Dataset. https://data.openei.org/files/5991/node_power_dist.csv
57 Metadata concerning each training run https://data.openei.org/files/5991/butter_e_metadata.csv.zip
57 Power data joined to run data, including extra columns for standardized energy data as described in the paper. https://data.openei.org/files/5991/runs_with_standardized_energy.csv.zip

Tags

  • energy-efficiency
  • network-structure
  • power-consumption
  • butter-e
  • deep-learning
  • empirical-machine-learning
  • benchmark
  • node-level
  • green-computing
  • neural-networks
  • energy-consumption
  • computational-science
  • power
  • model
  • energy-use
  • training-efficiency
  • butter
  • efficient
  • training
  • machine-learning
  • energy
  • empirical-deep-learning

Topics

Categories