can i use NPU to train a mmaction2 model???

[SherlockBULABULA]

can i use NPU to train a mmaction2 model?
in mmengine\device\utils.py there are some codes:

try:
    import torch_npu  # noqa: F401
    import torch_npu.npu.utils as npu_utils

    # Enable operator support for dynamic shape and
    # binary operator support on the NPU.
    npu_jit_compile = bool(os.getenv('NPUJITCompile', False))
    torch.npu.set_compile_mode(jit_compile=npu_jit_compile)
    IS_NPU_AVAILABLE = hasattr(torch, 'npu') and torch.npu.is_available()
except Exception:
    IS_NPU_AVAILABLE = False

is that means , i can choose NPU to tarin a model?