cann版本是8.3.RC1.alpha001，torch_npu版本是2.8.0rc1

报错内容：

Training: 31%|██▊ | 1837/6000 [1:10:28<2:31:16, 2.18s/it, loss=0.018, grad=0.93, lr=9.12e-05]

Training: 31%|██▊ | 1837/6000 [1:10:30<2:31:16, 2.18s/it, loss=0.028, grad=0.67, lr=9.12e-05]

Training: 31%|██▊ | 1838/6000 [1:10:30<2:32:09, 2.19s/it, loss=0.028, grad=0.67, lr=9.12e-05][rank1]: Traceback (most recent call last):

[rank1]: File "/root/anaconda3/envs/lerobot/lib/python3.10/runpy.py", line 196, in _run_module_as_main

[rank1]: return _run_code(code, main_globals, None,

[rank1]: File "/root/anaconda3/envs/lerobot/lib/python3.10/runpy.py", line 86, in _run_code

[rank1]: exec(code, run_globals)

[rank1]: File "/root/zzx/workspace/lerobot_ascend/src/lerobot/scripts/lerobot_train.py", line 448, in <module>

[rank1]: main()

[rank1]: File "/root/zzx/workspace/lerobot_ascend/src/lerobot/scripts/lerobot_train.py", line 444, in main

[rank1]: train()

[rank1]: File "/root/zzx/workspace/lerobot_ascend/src/lerobot/configs/parser.py", line 225, in wrapper_inner

[rank1]: response = fn(cfg, *args, **kwargs)

[rank1]: File "/root/zzx/workspace/lerobot_ascend/src/lerobot/scripts/lerobot_train.py", line 355, in train

[rank1]: train_tracker, output_dict = update_policy(

[rank1]: File "/root/zzx/workspace/lerobot_ascend/src/lerobot/scripts/lerobot_train.py", line 158, in update_policy

[rank1]: loss, output_dict = policy.forward(batch)

[rank1]: File "/root/anaconda3/envs/lerobot/lib/python3.10/site-packages/torch/nn/parallel/distributed.py", line 1648, in forward

[rank1]: else self._run_ddp_forward(*inputs, **kwargs)

[rank1]: File "/root/anaconda3/envs/lerobot/lib/python3.10/site-packages/torch/nn/parallel/distributed.py", line 1474, in _run_ddp_forward

[rank1]: return self.module(*inputs, **kwargs) # type: ignore[index]

[rank1]: File "/root/anaconda3/envs/lerobot/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl

[rank1]: return self._call_impl(*args, **kwargs)

[rank1]: File "/root/anaconda3/envs/lerobot/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl

[rank1]: return forward_call(*args, **kwargs)

[rank1]: File "/root/zzx/workspace/lerobot_ascend/src/lerobot/policies/smolvla/modeling_smolvla.py", line 326, in forward

[rank1]: losses = self.model.forward(images, img_masks, lang_tokens, lang_masks, state, actions, noise, time)

[rank1]: File "/root/zzx/workspace/lerobot_ascend/src/lerobot/policies/smolvla/modeling_smolvla.py", line 678, in forward

[rank1]: time = self.sample_time(actions.shape[0], actions.device)

[rank1]: File "/root/zzx/workspace/lerobot_ascend/src/lerobot/policies/smolvla/modeling_smolvla.py", line 529, in sample_time

[rank1]: time_beta = beta_dist.sample((bsize,)).to(device=device, dtype=torch.float32)

[rank1]: RuntimeError: ACL stream synchronize failed, error code:107020