MMGRec

This is our experiment codes for the paper:

MMGRec: Multimodal Generative Recommendation with Transformer Model

Environment settings

• Python 3.7
• Pytorch 1.7.0+cu101
• PyTorch Geometric 1.7.2
• Numpy 1.19.5

File specification

• data_load.py : loads the raw data.
• data_pro.py : processes the data further.
• src_input.py : obtains the historical interaction sequences of users.
• tgt_input.py : obtains the Rec-IDs of items.
• model_train.py : the training process of MMGRec.
• model_test.py : the testing process of MMGRec.

Usage

• Execution sequence

  The execution sequence of codes is as follows: data_load.py--->data_pro.py--->src_input.py--->tgt_input.py--->model_train.py--->model_test.py

• Execution results

  During the execution of file model_train.py, the epoch and training loss will be printed as the training process:

  Epoch: 0001 loss = 4.164487
  Epoch: 0002 loss = 3.460217
  Epoch: 0003 loss = 3.060792
  Epoch: 0004 loss = 2.914330
  ...
  

  File model_test.py should be executed after the training process, and the performance of MMGRec will be printed:

  R@10: 0.1269; NDCG@10: 0.0802