Deep Neural Network Prediction

Step 1. Genernate random parameters and Run them sequentially : $ python3 collect_data.py -gp -ep -pp -pl pooling -num 10 -shuffle -d 1080ti

Step 2. Data Timeline parser : $ python3 preprocess_data.py -pt -pl pooling -d 1080ti

Step 3. Combine all raw data : $ python3 preprocess_data.py -c -pl pooling -d 1080ti

Step 4. Split raw data to train and test data as performance prediction inputs $ python3 preprocess_data.py -sp -pl pooling -d 1080ti

Step 5. Train Model :

python3 train_model.py -ftf ./utils/Feature_Target/conv_pre.json -log2file 1 -e 1000 -st 400 -sg 0.5 -lf maple -n perfnetA -pd 1080ti -pl convolution -psl pre
    
python3 train_model.py -ftf ./utils/Feature_Target/conv_exe.json -log2file 1 -e 1000 -st 400 -sg 0.5 -lf maple -n perfnetA -pd 1080ti -pl convolution -psl exe
        
python3 train_model.py -ftf ./utils/Feature_Target/conv_post.json -log2file 1 -e 1000 -st 400 -sg 0.5 -lf maple -n perfnetA -pd 1080ti -pl convolution -psl post

Step 6. Generate model csv file :

python3 verify_model.py -gmc --model lenet -b 1

Step 7. Predict model csv file :

python3 verify_model.py -pdm -n perfnetA -d 1080ti -l -eva -lf maple --model lenet -b 1

Or you can use script to predict all models with all batches:

python3 predict_script.py -n perfnetA -d 1080ti -lf maple

Predict the latency time of the deep learning models

Related tags

Overview

Deep Neural Network Prediction

Owner

QAQ

SIR model parameter estimation using a novel algorithm for differentiated uniformization.

Implementation of a Transformer that Ponders, using the scheme from the PonderNet paper

Decorators for maximizing memory utilization with PyTorch & CUDA

Code for Quantifying Ignorance in Individual-Level Causal-Effect Estimates under Hidden Confounding

HNECV: Heterogeneous Network Embedding via Cloud model and Variational inference

Official Implementation of "LUNAR: Unifying Local Outlier Detection Methods via Graph Neural Networks"

End-to-end face detection, cropping, norm estimation, and landmark detection in a single onnx model

Weight initialization schemes for PyTorch nn.Modules

The repository contains source code and models to use PixelNet architecture used for various pixel-level tasks. More details can be accessed at .

The 2nd place solution of 2021 google landmark retrieval on kaggle.

Open-AI's DALL-E for large scale training in mesh-tensorflow.

shufflev2-yolov5：lighter, faster and easier to deploy

Official PyTorch implementation of RobustNet (CVPR 2021 Oral)

[ACMMM 2021 Oral] Enhanced Invertible Encoding for Learned Image Compression

Classification Modeling: Probability of Default

10x faster matrix and vector operations

The GitHub repository for the paper: “Time Series is a Special Sequence: Forecasting with Sample Convolution and Interaction“.

Pytorch implementation of "Training a 85.4% Top-1 Accuracy Vision Transformer with 56M Parameters on ImageNet"

SMCA replication There are no extra compiled components in SMCA DETR and package dependencies are minimal

This implements one of result networks from Large-scale evolution of image classifiers