Our experiments are conducted on a Ubuntu 64-Bit Linux workstation, having NVIDIA GeForce RTX 3090 GPUs with 24GB graphics memory. Conducting our experiments requires the following steps.
conda create --name <ENV-NAME> python=3.7
source activate <ENV-NAME>
conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
pip install -r requirements.txt
To conduct our experiments quickly, you need the following steps.
Two-task learning
python run_2task.py
Long-sequence learning
python run_ls.py
Extended analysis
#Robustness level
python run_extended_eps.py
#Model size
python run_extended_model_size.py
#Task Order
python run_extended_order.py
The file structure of directory is as follows:
.
├── CL-Robustness # The main code directory
│ ├── src
│ │ ├── data # The directory contains the dataset.
│ │ ├── framework # The directory contains the framework of continual learning.
│ │ ├── methods # The directory contains the codes of finetune.
│ │ ├── models # The directory contains the defined models.
│ │ ├── results # The directory contains the results.
│ │ ├── utilities # The directory contains some defined functions.
│ │ ├── config.init # The configuration file.
│ ├── main.py # The main code file.
│ ├── README.md # The readme file
│ ├── requirements.txt # The file specifies the required environments
│ ├── run_2task.py # The file to conduct two-task learning.
│ ├── run_extended_eps.py # The file to conduct extended experiments with different epsilon.
│ ├── run_extended_model_size.py # The file to conduct extended experiments with different model size.
│ ├── run_extended_order.py # The file to conduct extended experiments with different order.
│ ├── run_ls.py # The file to conduct long-sequence learning.
The results can be found in the directory "records":
├── CL-Robustness
│ ├── src
│ │ ├── results
│ │ │ ├── records