The official implementation of Doubly Stochastic Generative Arrivals Modeling
TODO:
- set appropriate server scheduling for multi server queue experiment (maybe a individual ipynb notebook).
- Verify results.
- erase the warning when running "python setup.py install"
- add args message for des_cpp
- CUDA support.
- Make the Poisson simulator layer into a PyTorch self defined layer
- Arrival epochs simulator
- Sample CIR process
- Run-through-queue
- Speed up multi server queue, C++ reimplement or multi-thread in python.
We recommend using conda environment.
conda create --name dswgan
conda install -c anaconda scipy -y
conda install -c conda-forge matplotlib -y
pip install progressbar2
conda install -c conda-forge colored -y
conda install pytorch -c pytorch -y
pip install geomloss
conda install -c anaconda seaborn -y
To build and install the C++ implementad descrete event simulation library (mainly for simulating multi-server queue), we assume you have installed the adequate C++ compiler. After that, activate the conda environment, then
conda install -c conda-forge pybind11 -y
cd core/des/des_cpp
python setup.py install
which will compile and install the discrete event simulation library to your conda environment.
Usage
python main.py --dataset cir
python main.py --dataset uniform
python main.py --dataset bimodal
python main.py --dataset bikeshare
python main.py --dataset callcenter
python main.py --dataset pgnorta
| Dataset | Verified code | Verified results |
|---|---|---|
| CIR | ✅ | |
| Uniform | ✅ | |
| Bimodal | ✅ | |
| Bikeshare | ✅ | |
| Callcenter | ✅ | |
| PGnorta | ✅ |