[Bugfix][LoRA][Patch] Fix the LoRA inference bug after upstream vLLM codebase changed (#2560)

### What this PR does / why we need it?
The mergence of the upstream
vllm-project/vllm#22592 caused a vllm-ascend
LoRA inference bug. The details are following:

According to
[torch_npu/npu/_stream_check.py](https://github.com/Ascend/pytorch/blob/863b9071cbdf47023c12c246e3efa9c6e2285fc6/torch_npu/npu/_stream_check.py#L74),
NPU device type tensors have attributes is_cuda=True and is_npu=True.
This causes that vLLM's apply_repetition_penalties function will run
into the branch of "if logits.is_cuda and logits.is_contiguous()" and
call the custom op implemented in CUDA, which is not compatible with
NPU.

### Does this PR introduce _any_ user-facing change?
No.

### How was this patch tested?
pytest -sv tests/e2e/singlecard/test_ilama_lora.py
pytest -sv tests/e2e/multicard/test_ilama_lora_tp2.py

- vLLM version: v0.10.1.1
- vLLM main:
vllm-project/vllm@fe8d7b6

---------

Signed-off-by: paulyu12 <paulyu0307@gmail.com>
Signed-off-by: paulyu12 <507435917@qq.com>
Co-authored-by: paulyu12 <paulyu0307@gmail.com>

Author: paulyu12

vllm_ascend/patch/worker/patch_common/__init__.py

@@ -17,4 +17,5 @@
 
 import vllm_ascend.patch.worker.patch_common.patch_distributed  # noqa
 import vllm_ascend.patch.worker.patch_common.patch_linear  # noqa
+import vllm_ascend.patch.worker.patch_common.patch_logits  # noqa
 import vllm_ascend.patch.worker.patch_common.patch_minicpm  # noqa

vllm_ascend/patch/worker/patch_common/patch_logits.py

@@ -0,0 +1,26 @@
+import torch
+import vllm
+from vllm._custom_ops import apply_repetition_penalties_torch
+
+
+def apply_repetition_penalties(logits: torch.Tensor, prompt_mask: torch.Tensor,
+                               output_mask: torch.Tensor,
+                               repetition_penalties: torch.Tensor) -> None:
+    """Apply repetition penalties to logits in-place.
+
+    Args:
+        logits: The logits tensor of shape [num_seqs, vocab_size].
+        prompt_mask: A boolean tensor indicating which tokens appear in the prompt.
+        output_mask: A boolean tensor indicating which tokens appear in the output.
+        repetition_penalties: The repetition penalties of shape (num_seqs, ).
+    """
+    apply_repetition_penalties_torch(logits, prompt_mask, output_mask,
+                                     repetition_penalties)
+
+
+# NPU device type tensors have attributes is_cuda=True and is_npu=True, according to its implementation in
+# https://github.com/Ascend/pytorch/blob/863b9071cbdf47023c12c246e3efa9c6e2285fc6/torch_npu/npu/_stream_check.py#L74
+# This causes that vLLM's apply_repetition_penalties function will run into the branch of "if logits.is_cuda" and
+# call the custom op implemented in CUDA, which is not compatible with NPU.
+# Reference: https://github.com/vllm-project/vllm/blob/f66673a39d9f364194c249f28098cad8a5584ccb/vllm/_custom_ops.py#L314
+vllm._custom_ops.apply_repetition_penalties = apply_repetition_penalties