Update configuration_quasrav4.py

configuration_quasrav4.py

@@ -29,6 +29,7 @@ class InfinityFormerConfig(PretrainedConfig):
         self.kernel_epsilon = kwargs.pop("kernel_epsilon", 0.1)
         self.use_gating = kwargs.pop("use_gating", True)
         self.gate_init_bias = kwargs.pop("gate_init_bias", -2.0)
+        self.use_memory_attention = kwargs.pop("use_memory_attention", False)
         self.use_gradient_checkpointing = kwargs.pop("use_gradient_checkpointing", False)
 
         # The `use_return_dict` is a read-only property that depends on `return_dict`.