You need to modify the dataset path and the pretrained language model path in config_regression.json to your own. If you don't intend to train the model, you can load the pre-trained weights we provide. In this case, you also need to modify the weight download URL in run.py and comment out the training-related code.
Download the MOSI and MOSEI pkl file https://drive.google.com/drive/folders/1_u1Vt0_4g0RLoQbdslBwAdMslEdW1avI?usp=sharing. Put it under the "./dataset" directory.
Download the Bert language model files https://storage.googleapis.com/bert_models/2018_10_18/uncased_L-12_H-768_A-12.zip.
If you do not want to train the model, you can load the pre-trained model weights. The training weights obtained on MOSI and MOSEI are available at the following link: https://drive.google.com/drive/folders/1hJ8cD_piHwOGe5FY8s5BmOvx8PnaqPlo?usp=drive_link.
''' python test.py '''
Note: To get results close to those in our paper, you can set the seed in args to 1111. The experimental results of this paper are obtained on the Linux system.