卡在 loading model.layers.0.self_attn.k_norm.weight to cuda request to allocate mask for invalid number: Invalid argument

CPU ：AMD9950X
GPU ： 4090 * 2
操作系统 wsl ubuntu 22.04
运行模型Qwen3-30B-A22B 卡在卡在 loading model.layers.0.self_attn.k_norm.weight to cuda request to allocate mask for invalid number: Invalid argument

Injecting cache as default
Injecting lm_head as ktransformers.operators.linear . KTransformersLinear
loading model.embed_tokens.weight to cpu
loading model.layers.0.self_attn.q_norm.weight to cuda
loading model.layers.0.self_attn.k_norm.weight to cuda
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
request to allocate mask for invalid number: Invalid argument
^CTraceback (most recent call last):
File "/usr/local/ktransformers/ktransformers/server/main.py", line 122, in
main()
File "/usr/local/ktransformers/ktransformers/server/main.py", line 109, in main
create_interface(config=cfg, default_args=cfg)
File "/root/miniforge3/envs/ktransforms/lib/python3.12/site-packages/ktransformers/server/utils/create_interface.py", line 30, in create_interface
GlobalInterface.interface = BackendInterface(default_args)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/miniforge3/envs/ktransforms/lib/python3.12/site-packages/ktransformers/server/backend/interfaces/balance_serve.py", line 342, in __init__
kvcache_event.wait()
File "/root/miniforge3/envs/ktransforms/lib/python3.12/multiprocessing/synchronize.py", line 356, in wait
self._cond.wait(timeout)
File "/root/miniforge3/envs/ktransforms/lib/python3.12/multiprocessing/synchronize.py", line 268, in wait
return self._wait_semaphore.acquire(True, timeout)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
KeyboardInterrupt

请问各位大佬这个什么原因啊 怎么解决

Reproduce the Qwen3-30B-A22B loading failure in the reported WSL Ubuntu 22.04 setup with two RTX 4090 GPUs. Start with ktransformers/server/main.py and the balance_serve.py path around BackendInterface and kvcache_event.wait(); determine why allocation repeatedly reports an invalid mask, then verify that model loading completes without the repeated error.

Written by the indexing model from the issue text.