update the CUDA to accelerate model inference if there is any.

tapex/model_interface.py

@@ -4,6 +4,7 @@
 import logging
 from typing import Dict, List
 
+import torch.cuda
 from fairseq.models.bart import BARTModel
 
 from tapex.processor import get_default_processor
@@ -19,6 +20,8 @@ class TAPEXModelInterface:
     def __init__(self, resource_dir, checkpoint_name, table_processor=None):
         self.model = BARTModel.from_pretrained(model_name_or_path=resource_dir,
                                                checkpoint_file=checkpoint_name)
+        if torch.cuda.is_available():
+            self.model.cuda()
         self.model.eval()
         if table_processor is not None:
             self.tab_processor = table_processor