fix orpo chosen-nll loss (#2502)

huggingface · Dec 19, 2024 · 88ad1a0 · 88ad1a0
1 parent 9908dda
commit 88ad1a0
Showing 1 changed file with 8 additions and 10 deletions.
diff --git a/trl/trainer/orpo_trainer.py b/trl/trainer/orpo_trainer.py
@@ -774,18 +774,12 @@ def cross_entropy_loss(logits, labels):
             loss = loss_fct(logits, labels)
             return loss
 
-        if self.is_encoder_decoder:
-            labels = concatenated_batch["concatenated_labels"].clone()
-        else:
-            labels = concatenated_batch["concatenated_input_ids"].clone()
-            attention_mask = concatenated_batch["concatenated_attention_mask"]
-            labels = torch.where(attention_mask == 1, labels, self.label_pad_token_id)
-
+        labels = concatenated_batch["concatenated_labels"].clone()
         chosen_nll_loss = cross_entropy_loss(all_logits[:len_chosen], labels[:len_chosen])
 
         all_logps = self.get_batch_logps(
             all_logits,
-            concatenated_batch["concatenated_labels"],
+            labels,
             average_log_prob=True,
             is_encoder_decoder=self.is_encoder_decoder,
             label_pad_token_id=self.label_pad_token_id,
@@ -794,8 +788,12 @@ def cross_entropy_loss(logits, labels):
         chosen_logps = all_logps[:len_chosen]
         rejected_logps = all_logps[len_chosen:]
 
-        chosen_logits = all_logits[:len_chosen]
-        rejected_logits = all_logits[len_chosen:]
+        if not self.is_encoder_decoder:
+            chosen_logits = all_logits[:len_chosen, :-1, :]
+            rejected_logits = all_logits[len_chosen:, :-1, :]
+        else:
+            chosen_logits = all_logits[:len_chosen]
+            rejected_logits = all_logits[len_chosen:]
 
         if self.aux_loss_enabled:
             return (chosen_logps, rejected_logps, chosen_logits, rejected_logits, chosen_nll_loss, outputs.aux_loss)