diff --git a/vllm/lora/layers.py b/vllm/lora/layers.py
index aa01e9fb77af..015dbb8bd0d5 100644
--- a/vllm/lora/layers.py
+++ b/vllm/lora/layers.py
@@ -1250,6 +1250,8 @@ def _get_logits(
                            nan=float("-inf"),
                            posinf=float("inf"),
                            neginf=float("-inf")))
+        if is_hpu():
+            lora_logits = lora_logits[:logits.shape[0], :]
         logits[:,
                self.base_layer.org_vocab_size:self.base_layer.org_vocab_size +
                lora_logits.shape[1]] = lora_logits
diff --git a/vllm/lora/models.py b/vllm/lora/models.py
index e8d39591cb17..f819e3816fd3 100644
--- a/vllm/lora/models.py
+++ b/vllm/lora/models.py
@@ -24,7 +24,7 @@
 from vllm.lora.utils import (from_layer, from_layer_logits_processor,
                              parse_fine_tuned_lora_name, replace_submodule)
 from vllm.model_executor.models.interfaces import SupportsLoRA
-from vllm.utils import get_device, is_pin_memory_available
+from vllm.utils import get_device, is_hpu, is_pin_memory_available
 
 logger = init_logger(__name__)
 
@@ -829,6 +829,8 @@ def create_lora_manager(
     """Create a LoRA adapter for a given model."""
     if not hasattr(model, "supported_lora_modules"):
         raise ValueError(f"Model {type(model)} is not supported for LoRA.")
+    if is_hpu():
+        max_num_batched_tokens = 3 * max_num_batched_tokens
     lora_manager = lora_manager_cls(
         model=model,
         max_num_seqs=max_num_seqs,
diff --git a/vllm/worker/habana_model_runner.py b/vllm/worker/habana_model_runner.py
index 161364ce05f9..685e9c0140d8 100644
--- a/vllm/worker/habana_model_runner.py
+++ b/vllm/worker/habana_model_runner.py
@@ -1937,14 +1937,6 @@ def execute_model(
             )
 
         if self.lora_config:
-            from vllm.lora.layers import VocabParallelEmbeddingWithLoRA
-            modules = unwrap_model(self.model.model)
-            for module in modules:
-                if isinstance(module, VocabParallelEmbeddingWithLoRA):
-                    for i in range(0, len(module.indices_len)):
-                        module.indices_len[
-                            i] = sampling_metadata.selected_token_indices.numel(
-                            )
             lora_logits_mask: torch.Tensor = model_input.lora_logits_mask
             LoraMask.setLoraMask(
                 lora_logits_mask.index_select(