RuntimeError: CUDA error: no kernel image is available for execution on the device, which about xformers

Question

does anyone can help me solve this problem?

File "/opt/conda/envs/vgen/lib/python3.8/site-packages/fairscale/nn/checkpoint/checkpoint_activations.py", line 170, in _checkpointed_forward
    return original_forward(module, *args, **kwargs)
File "/workspace/tools/modules/unet/util.py", line 1261, in forward
    x = block(x)
File "/opt/conda/envs/vgen/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
    return forward_call(*input, **kwargs)
File "/workspace/tools/modules/unet/util.py", line 701, in forward
    x = self.attn1(self.norm1(x), context=context if self.disable_self_attn else None) + x
File "/opt/conda/envs/vgen/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1130, in _call_impl
    return forward_call(*input, **kwargs)
File "/workspace/tools/modules/unet/util.py", line 254, in forward
    out = xformers.ops.memory_efficient_attention(
File "/opt/conda/envs/vgen/lib/python3.8/site-packages/xformers/ops.py", line 574, in memory_efficient_attention
    return op.forward_no_grad(
  File "/opt/conda/envs/vgen/lib/python3.8/site-packages/xformers/ops.py", line 189, in forward_no_grad
    return cls.FORWARD_OPERATOR(
File "/opt/conda/envs/vgen/lib/python3.8/site-packages/torch/_ops.py", line 143, in __call__
    return self._op(*args, **kwargs or {})
RuntimeError: CUDA error: no kernel image is available for execution on the device

I use pytorch docker: enter image description here

and related pip list: enter image description here

and the xformers is: enter image description here

The github repository is: VGen

and it's commands:

conda create -n vgen python=3.8
conda activate vgen
pip install torch==1.12.0+cu113 torchvision==0.13.0+cu113 torchaudio==0.12.0 --extra-index-url https://download.pytorch.org/whl/cu113
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple

I follow this suggestion:RuntimeError: CUDA error: no kernel image is available for execution on the device (rastervision)

But it doesn't work.