Update modeling_gpt_refact.py

modeling_gpt_refact.py

@@ -503,6 +503,18 @@ class GPTRefactForCausalLM(GPTRefactPreTrainedModel):
 
         # Initialize weights and apply final processing
         self.post_init()
+        
+        # gradient checkpointing support for lower versions of transformers
+        import transformers
+        from packaging import version
+
+        def _set_gradient_checkpointing(module, enable=False):
+            if isinstance(module, GPTRefactModel):
+                module.gradient_checkpointing = enable
+
+        v = version.parse(transformers.__version__)
+        if v.major <= 4 and v.minor < 35:
+            self._set_gradient_checkpointing = _set_gradient_checkpointing
 
     def prepare_inputs_for_generation(self, input_ids, past_key_values=None, inputs_embeds=None, **kwargs):
         if inputs_embeds is not None and past_key_values is None: