manual rebase with main

Author: gante

src/transformers/models/llama/modeling_llama.py

@@ -871,7 +871,6 @@ def forward(
             if not isinstance(past_key_values, Cache):
                 past_key_values = self.from_legacy_cache(past_key_values)
             past_key_values_length = past_key_values.get_seq_length()
-            seq_length_with_past += past_key_values_length
 
         if position_ids is None:
             device = input_ids.device if input_ids is not None else inputs_embeds.device
@@ -917,7 +916,7 @@ def forward(
                     hidden_states,
                     attention_mask,
                     position_ids,
-                    past_key_value,
+                    past_key_values,
                     output_attentions,
                     use_cache,
                 )