we propose CKD-MDSR, which stands for Curriculum-scheduled Knowledge Distillation from multiple pre-trained teachers for Multi-Domain Sequential Recommendation. Aiming to selectively integrate the informative knowledge distilled from various PRMs into a student model.
recbole>=1.1.1
python>=3.9.7
cudatoolkit>=11.3.1
pytorch>=1.11.0
numpy>=1.21.6
cupy>=11.6.0
tqdm>=4.64.1
If you have any question, you can raise an issue or email Wenqi Sun ([email protected]). We will reply you soon.