Upload 7 files

modeling_indictrans.py

@@ -61,6 +61,11 @@ def shift_tokens_right(input_ids: torch.Tensor, pad_token_id: int, decoder_start
 
 
 def prepare_decoder_input_ids_label(decoder_input_ids, decoder_attention_mask):
+    labels = decoder_input_ids[:, 1:]
+    
+    labels_mask = labels == 1
+    labels[labels_mask] = -100
+    
     mask = (decoder_input_ids == eos_token_id)
     decoder_input_ids[mask] = 1
     decoder_attention_mask[mask] = 0