update src

README.md

@@ -121,7 +121,7 @@ The result for dev and the test set below. We recommend using 0.25 as cut-off sc
 | Model             | Split	| Q only | Q + A | Veracity @ 0.2 | @ 0.25 | @ 0.3 |
 |-------------------|-------|--------|-------|----------------|--------|-------|
 | AVeriTeC-BLOOM-7b | dev	|  0.24  | 0.19  | 	    0.19      |  0.09  |  0.05 |
-| AVeriTeC-BLOOM-7b | test	|    	|   	|  	|  	| 	|
+| AVeriTeC-BLOOM-7b | test	|  0.25  | 0.18	 |  	0.18      |  0.11  |  0.06 |
 
 ## Citation
 If you find AVeriTeC useful for your research and applications, please cite us using this BibTeX:

src/prediction/evaluate_veracity.py

@@ -264,7 +264,7 @@ if __name__ == "__main__":
     parser.add_argument(
         "-i",
         "--prediction_file",
-        default="data_store/dev_veracity.json",
+        default="data_store/dev_veracity_prediction.json",
         help="Json file with claim, evidence, and veracity prediction.",
     )
     parser.add_argument(

src/prediction/veracity_prediction.py

@@ -26,7 +26,7 @@ class SequenceClassificationDataLoader(pl.LightningDataModule):
     def tokenize_strings(
         self,
         source_sentences,
-        max_length=512,
+        max_length=400,
         pad_to_max_length=False,
         return_tensors="pt",
     ):
@@ -72,7 +72,7 @@ if __name__ == "__main__":
     parser.add_argument(
         "-o",
         "--output_file",
-        default="data_store/dev_veracity.json",
+        default="data_store/dev_veracity_prediction.json",
         help="Json file with the veracity predictions.",
     )
     parser.add_argument(