Requirements

numpy==1.19.5
pandas==1.3.3
scikit_learn==1.0.1
torch==1.9.0
torch_geometric==2.0.1
tqdm==4.62.3
transformers==4.11.3

Running procedure

1. construct_graph.py - Constructs the necessary semantic embeddings and graph neighborhood
2. dataloader.py - Builds the text and graph dataloaders.
3. train.py - To train the model on the entire dataset.
4. finetune.py - To finetune the model on marketplace-specific dataset.
5. evaluate.py - To evaluate the model on datasets.

Arguments:

To find model arguments, run:

python <script> --help

Code details

construct_graph.py - Constructs the necessary semantic embeddings and graph neighborhood
dataloader.py - Builds the text and graph dataloaders.
train.py - To train the model on the entire dataset.
model.py - Defines the SMLM (Classifier) model. 
finetune.py - To finetune the model on marketplace-specific dataset.
evaluate.py - To evaluate the model on datasets.