Fix NandiModel.forward: add cache_position for compatibility with released transformers 5.3.0

modeling_nandi.py

@@ -354,6 +354,7 @@ class NandiModel(NandiPreTrainedModel):
         position_ids: torch.LongTensor | None = None,
         past_key_values: Cache | None = None,
         inputs_embeds: torch.FloatTensor | None = None,
+        cache_position: torch.LongTensor | None = None,
         use_cache: bool | None = None,
         **kwargs: Unpack[TransformersKwargs],
     ) -> BaseModelOutputWithPast:
@@ -369,19 +370,20 @@ class NandiModel(NandiPreTrainedModel):
         repeats = self.config.layer_sharing_repeats if self.config.layer_sharing else 1
 
         if use_cache and past_key_values is None:
-            # Use lazy DynamicCache (no config) so it grows to accommodate
-            # num_hidden_layers * repeats virtual slots for layer-sharing.
             past_key_values = DynamicCache()
 
-        if position_ids is None:
+        if cache_position is None:
             past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
-            position_ids = torch.arange(inputs_embeds.shape[1], device=inputs_embeds.device) + past_seen_tokens
-            position_ids = position_ids.unsqueeze(0)
+            cache_position = torch.arange(inputs_embeds.shape[1], device=inputs_embeds.device) + past_seen_tokens
+
+        if position_ids is None:
+            position_ids = cache_position.unsqueeze(0)
 
         causal_mask = create_causal_mask(
             config=self.config,
             inputs_embeds=inputs_embeds,
             attention_mask=attention_mask,
+            cache_position=cache_position,
             past_key_values=past_key_values,
             position_ids=position_ids,
         )
@@ -391,8 +393,6 @@ class NandiModel(NandiPreTrainedModel):
 
         for decoder_layer in self.layers[: self.config.num_hidden_layers]:
             for repeat_idx in range(repeats):
-                # Each repeat gets its own virtual cache slots offset by num_hidden_layers,
-                # so repeat 0 uses slots 0..N-1 and repeat 1 uses slots N..2N-1, etc.
                 repeat_cache = (
                     _VirtualLayerCache(past_key_values, repeat_idx * self.config.num_hidden_layers)
                     if (past_key_values is not None and repeat_idx > 0)
@@ -405,6 +405,7 @@ class NandiModel(NandiPreTrainedModel):
                     position_ids=position_ids,
                     past_key_values=repeat_cache,
                     use_cache=use_cache,
+                    cache_position=cache_position,
                     **kwargs,
                 )