https://github.com/xuhwang/corr_www
Cooperation of Retriever and Ranker Framework.
https://github.com/xuhwang/corr_www
knowledge-distillation negative-sampling ranking recommender-system retrieval
Last synced: 8 months ago
JSON representation
Cooperation of Retriever and Ranker Framework.
- Host: GitHub
- URL: https://github.com/xuhwang/corr_www
- Owner: XuHwang
- Created: 2022-10-13T12:21:15.000Z (over 3 years ago)
- Default Branch: master
- Last Pushed: 2022-10-13T12:27:25.000Z (over 3 years ago)
- Last Synced: 2025-09-12T10:21:56.301Z (10 months ago)
- Topics: knowledge-distillation, negative-sampling, ranking, recommender-system, retrieval
- Language: Python
- Homepage:
- Size: 1.72 MB
- Stars: 7
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: Readme.md
Awesome Lists containing this project
README
# CoRR
Here is the code for Cooperative Retriever and Ranker in Deep Recommenders based on recommendation library [RecStudio](https://github.com/ustcml/RecStudio)
### Dataset
With RecStudio, the dataset can be downloaded automatically by specifying dataset name.
### Run
To run CoRR algorithm, you should run:
```bash
python run.py -m CoRR -d amazon-electronics --batch_size 512
```
If you just want to have a try, a tiny dataset is recommended: ml-100k.
For general recommendation, you should run:
```bash
python run.py -m CoRRMF -d amazon-electronics --batch_size 1024
```
The default retriever and ranker are [SASRec, DIN] for sequential recommendation and [MF+DeepFM] for general recommendation. If you want to specify retriever and ranker, run like this:
```bash
python run.py -m CoRR -d amazon-electronics --retriever Caser --ranker BST
```
The default number of negatives is 20, you can specify it with arguments `--num_neg`, i.e. `--num_neg 100`.