Update configuration_deepseekocr.py

configuration_deepseekocr.py

@@ -61,6 +61,8 @@ class DeepseekOCRConfig(PretrainedConfig):
         self.num_key_value_heads = num_key_value_heads
         self.max_position_embeddings = max_position_embeddings
         self.moe_intermediate_size = moe_intermediate_size
+        self.attention_dropout = kwargs.get("attention_dropout", 0.0)
+
 
         # Mixture of Experts & attention params
         self.n_group = n_group