In this repository we provide access to the code used to reproduce the results from our work "Fast Incremental Learning by Transfer Learning and Hierarchical Sequencing".
Our experiments can be reproduced in any operating system that supports the following dependencies:
- Python 3.6.9
- CUDA 11.0 (optional, only for faster training/testing times)
-
Download our code repository:
git clone https://github.com/capo-urjc/HILAND.git
-
Move into the cloned project:
cd HILAND
-
Create a python virtual environment:
python -m venv venv
-
Activate the newly created virtual environment:
-
Linux:
source venv/bin/activate
-
Windows:
venv\Scripts\activate.bat
-
-
Upgrade
pip
to prevent errors while installing newer packages:pip install --upgrade pip
-
Install our basic requirements:
pip install -r requirements.txt
-
Install the required PyTorch version:
-
If using a CUDA 11.0 enabled GPU:
pip install torch==1.7.1+cu110 torchvision==0.8.2+cu110 torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html
-
Otherwise, install the CPU version (results may differ):
pip install torch==1.7.1+cpu torchvision==0.8.2+cpu torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html
-
To run our experiments we provide 20 configuration files in exp1_5/exp_config and exp6_7/exp_config. To run our experiments use the following commands:
-
From a terminal window in the project root directory, activate the virtual environment created during the installation process:
-
Linux:
source venv/bin/activate
-
Windows:
venv\Scripts\activate.bat
-
-
Run our experiments:
-
Linux:
./run_experiments.sh
(usechmod +x run_experiments.sh
if you encounter permission issues when running the previous command) -
Windows:
run_experiments.bat
-
-
Check the results in the CSV files in the newly created
exp_out
folder.