Divide-and-Conquer: Prompt-based Distribution Learning for Multimodal Sentiment Analysis

This is the code repository for the paper: "Divide-and-Conquer: Prompt-based Distribution Learning for Multimodal Sentiment Analysis"

Note

The code has been refactored. Please raise any issues.

Usage

1. Prepare the dataset and pre-trained RoBERTa. For dataset , you can download the processed data from Self-MM. For RoBERTa model, you can download robert-large and chinese-roberta-wwm-ext from Hugging Face.

2. Clone this repo and install requirements.

   git clone https://github.com/Dongri404/PDL.git
   cd PDL
   conda create --name pdl python=3.8
   source activate pdl
   pip install -r requirements.txt

3. Update path. Modify the config/config_tune.py and config/config_regression.py to update dataset paths . Update the pretrained model path in models/subNets/BertTextEncoder.py.

4. Run

   python -u run.py --datasetNames sims mosi mosei