search

enyac-group / NeuralPower

The code for paper: Neuralpower: Predict and deploy energy-efficient convolutional neural networks
Apache License 2.0
21 stars 7 forks source link

Sample data input from the data collection step #1

Open Crefeda opened 5 years ago

Crefeda commented 5 years ago

Hi,

I would like to use the parser_raw_data.py to parse the data from my own data collection process. Could you please provide a format of the raw data that goes into the function - parser_raw_data.py? Specifically the input_file to this function: parse_results(input_file, output_init='res'): Thanks

caiermao commented 5 years ago

Hi,

I am sorry for the delay in the response. The input raw data is actually generated from the scripts of the profiling from the Paleo. It lists the layer type, the key parameters for the layer, and the collected runtime/power for the layer. It then put same category of layer data into one file for further modeling and prediction purpose.

One can always use my modified Paleo (in this repo) to collect power/runtime data, and use this script to parse the results. Or one can write another parser to collect the profiling data one have.

Bests, Ermao