fix wrong filling chatglm position_ids input

Author: eaidova

optimum/intel/openvino/modeling_decoder.py

@@ -16,7 +16,7 @@
 import os
 from pathlib import Path
 from tempfile import TemporaryDirectory
-from typing import TYPE_CHECKING, Callable, Dict, List, Optional, Tuple, Union
+from typing import TYPE_CHECKING, Any, Callable, Dict, List, Optional, Tuple, Union
 
 import numpy as np
 import openvino
@@ -31,7 +31,7 @@
 from transformers.generation.logits_process import LogitsProcessorList
 from transformers.generation.stopping_criteria import StoppingCriteriaList
 from transformers.generation.utils import GenerateOutput, GenerationMode
-from transformers.modeling_outputs import CausalLMOutputWithPast
+from transformers.modeling_outputs import CausalLMOutputWithPast, ModelOutput
 
 from optimum.utils.normalized_config import NormalizedConfigManager
 
@@ -604,6 +604,24 @@ def prepare_inputs_for_generation(self, input_ids, past_key_values=None, **kwarg
 
         return model_inputs
 
+    def _update_model_kwargs_for_generation(
+        self,
+        outputs: ModelOutput,
+        model_kwargs: Dict[str, Any],
+        is_encoder_decoder: bool = False,
+        **kwargs,
+    ) -> Dict[str, Any]:
+        model_kwargs = super()._update_model_kwargs_for_generation(
+            outputs=outputs, model_kwargs=model_kwargs, is_encoder_decoder=is_encoder_decoder, **kwargs
+        )
+
+        if "position_ids" in model_kwargs:
+            position_ids = model_kwargs["position_ids"]
+            new_position_id = position_ids[..., -1:].clone()
+            new_position_id += 1
+            model_kwargs["position_ids"] = torch.cat([position_ids, new_position_id], dim=-1)
+        return model_kwargs
+
     def _expand_outputs_for_generation(self, indicies, logits: torch.Tensor, past_key_values: Tuple):
         batch_size = logits.shape[0]
         if indicies.shape[0] != 1: