The functional simulator models different components of matrix operations execution with ReRam backend.
Requirement | Version |
---|---|
Anaconda | 4.5.11 |
Python | 3.6.6 or higher |
Tensorflow-gpu | tested on 1.10.0 |
conda create --name tf-gpu python=3.6
source activate tf-gpu
conda install -c anaconda tensorflow-gpu
Download CIFAR-100
dataset and convert it to tfrecords files (train and test sets).
Data Loader data_loader.py
in repo has been tested for CIFAR-100 dataset, should be extensible to other datasets with changing the parameters such as number of classes, training and testing example etc.
The training and testing scripts use vgg16 models by default - ``vgg16_puma.py```.
Note: this version runs on 1 GPU only.
For training with PANTHER operations use ifpanther=True
.
python train_puma.py --dataset=<my_path>
python test.py --dataset=<my_path>
Tensorboard helps in visualization of several statistics collected during the training run (accuracy, loss etc).
tensorboard --logdir=<my_name>:<my_logpath>
Uncomment Lines 185-187, Line 194
in train_puma.py (and comment Lines 188 and 195
) to dump related metadata during training.
Launch tensorboard, choose run-step on left-pane, choose compute time or memory from the checklist.
Aayush Ankit (Purdue University), Sai Rahul Chalamalasetti (Hewlett Packard Labs)