Fix idefics cache (#31377)

* fix idefics cache

* fix tests

src/transformers/models/idefics2/modeling_idefics2.py

@@ -1593,10 +1593,11 @@ class Idefics2Model(Idefics2PreTrainedModel):
 
         past_seen_tokens = 0
         return_legacy_cache = False
-        if use_cache and not isinstance(past_key_values, Cache):  # kept for BC (non `Cache` `past_key_values` inputs)
+        if use_cache:
+            if not isinstance(past_key_values, Cache):  # kept for BC (non `Cache` `past_key_values` inputs)
                 return_legacy_cache = True
                 past_key_values = DynamicCache.from_legacy_cache(past_key_values)
-            past_seen_tokens = past_key_values.get_usable_length(seq_length)
+            past_seen_tokens = past_key_values.get_seq_length()
 
         if inputs_embeds is not None and input_ids is None and past_seen_tokens == 0:
             raise ValueError("When first calling the model, if input_embeds are passed, input_ids should not be None.")
@@ -1669,7 +1670,7 @@ class Idefics2Model(Idefics2PreTrainedModel):
             return_dict=return_dict,
         )
 
-        if return_legacy_cache:
+        if return_legacy_cache and use_cache:
             outputs.past_key_values = outputs.past_key_values.to_legacy_cache()
 
         if not return_dict:
@@ -1880,8 +1881,7 @@ class Idefics2ForConditionalGeneration(Idefics2PreTrainedModel):
         # Omit tokens covered by past_key_values
         if past_key_values is not None:
             # Past key values are always initialized with a `Cache` object -> no need for if-else anymore
-            cache_length = past_key_values.get_seq_length()
-            past_length = past_key_values.seen_tokens
+            past_length = past_key_values.get_seq_length()
             max_cache_length = past_key_values.get_max_length()
 
             # Keep only the unprocessed tokens:
@@ -1900,7 +1900,7 @@ class Idefics2ForConditionalGeneration(Idefics2PreTrainedModel):
             if (
                 max_cache_length is not None
                 and attention_mask is not None
-                and cache_length + input_ids.shape[1] > max_cache_length
+                and past_length + input_ids.shape[1] > max_cache_length
             ):
                 attention_mask = attention_mask[:, -max_cache_length:]