Skip to content

DreamCatcher-cai/PeerCollaboration

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

30 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

PeerCollaboration

Layer-wise cooperation

Execute example:

CUDA_VISIBLE_DEVICES=0 python3 -u SASRec_LW.py --datapath ml20m_removecold5_seq.csv --savedir ml20m/ --max_len 100 --num 2 --i 1 --a 30 --seed 10 --cos --difflr --cooperation_type 'cooperation_type' > 1.log & CUDA_VISIBLE_DEVICES=1 python3 -u SASRec_LW.py --datapath ml20m_removecold5_seq.csv --savedir ml20m/ --max_len 100 --num 2 --i 2 --a 30 --seed 11 --cos --difflr --cooperation_type 'cooperation_type' > 2.log  

The import configuration:

--cooperation_type: This is used to control which cooperation type is used. There are 5 parameters in SASRec_layercooperation.py:  
alllayer_entropy:All layers are used by layer-wise cooperation with entropy criterion.
alllayer_eL1norm:All layers are used by layer-wise cooperation with L1-norm criterion.
onlyembed:Only the embedding layer uses layer-wise cooperation with entropy criterion.
onlymiddle:All layers exception the embedding and final layer use layer-wise cooperation with entropy criterion.  
onlyfinal: Only the final layer uses layer-wise cooperation with entropy criterion.

Parameter-wise cooperation

Execute example:

CUDA_VISIBLE_DEVICES=0 python3 -u SASRec_PW.py --datapath ml20m_removecold5_seq.csv --savedir ml20m/ --max_len 100 --num 2 --i 1 --percent 50 --seed 10 --cos --difflr > 1.log & CUDA_VISIBLE_DEVICES=1 python3 -u SASRec_PW.py --datapath ml20m_removecold5_seq.csv --savedir ml20m/ --max_len 100 --num 2 --i 2 --percent 50 --seed 11 --cos --difflr > 2.log  

you can download a large sequential dataset of movielen-20m that has been pre-processed: https://drive.google.com/drive/folders/1TYtwwQruNcdDPQymsEgNRraXtjMf9jdl?usp=sharing

References

https://github.com/kang205/SASRec code

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%