fix-get_max_length-deprecation

https://huggingface.co./microsoft/Phi-3-vision-128k-instruct/discussions/69 , https://github.com/huggingface/transformers/issues/36071#issuecomment-2642222305

get_max_length deprecated with transformers 4.49

modeling_pharia.py

@@ -606,7 +606,7 @@ class PhariaModel(PhariaPreTrainedModel):
         min_dtype = torch.finfo(dtype).min
         sequence_length = input_tensor.shape[1]
         if using_static_cache:
-            target_length = past_key_values.get_max_length()
+            target_length = past_key_values.get_max_cache_shape()
         else:
             target_length = (
                 attention_mask.shape[-1]
@@ -812,9 +812,9 @@ class PhariaForCausalLM(PhariaPreTrainedModel):
                 )
                 max_cache_length = (
                     torch.tensor(
-                        past_key_values.get_max_length(), device=input_ids.device
+                        past_key_values.get_max_cache_shape(), device=input_ids.device
                     )
-                    if past_key_values.get_max_length() is not None
+                    if past_key_values.get_max_cache_shape() is not None
                     else None
                 )
                 cache_length = (