Fix the following runtime error with --no-use_cuda_graph option

Traceback (most recent call last):
  File "/home/aubrey/miniforge3/envs/kt/lib/python3.11/multiprocessing/process.py", line 314, in _bootstrap
    self.run()
  File "/home/aubrey/miniforge3/envs/kt/lib/python3.11/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/home/aubrey/miniforge3/envs/kt/lib/python3.11/site-packages/ktransformers/server/backend/interfaces/balance_serve.py", line 282, in run_engine
    engine.loop()
  File "/home/aubrey/miniforge3/envs/kt/lib/python3.11/site-packages/ktransformers/server/backend/interfaces/balance_serve.py", line 234, in loop
    self.model_runner.run(self.batch, self.query_manager)
  File "/home/aubrey/miniforge3/envs/kt/lib/python3.11/site-packages/ktransformers/server/balance_serve/inference/model_runner.py", line 220, in run
    self.output.logits[0] = self.output.logits[0][self.input[cuda_graph_idx].minibatch.logits_start]
                            ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: CUDA error: an illegal memory access was encountered
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.

VLinearMarlin: padding to input.shape[0] to avoid CUDA error

Update version

Fix NaN bug

add XPU support for qwen3moe local chat

- Add Dockerfile.xpu for oneAPI-based container
- Create Docker_xpu.md with usage instructions
- Update xpu.md to include Docker guide

Add Dockerfile and usage guide for XPU support

* display the unavailable torch device on error

* Raise exception on device error

---------

Signed-off-by: Emmanuel Ferdman <[email protected]>