DeepSpeed: Cuda 11 cannot be supported

I wanted to run Deepdpeed on RTX 3090, cudA 11 only on 3090, and in your docker release I updated Pytorch version to 1.7.0 and ran an error:

building GPT2 model ...
Traceback (most recent call last):
  File "pretrain_gpt2.py", line 711, in <module>
    main()
  File "pretrain_gpt2.py", line 659, in main
    model, optimizer, lr_scheduler = setup_model_and_optimizer(args)
  File "pretrain_gpt2.py", line 158, in setup_model_and_optimizer
    model = get_model(args)
  File "pretrain_gpt2.py", line 69, in get_model
    parallel_output=True)
  File "/data/Megatron-LM/model/gpt2_modeling.py", line 81, in __init__
    checkpoint_num_layers)
  File "/data/Megatron-LM/mpu/transformer.py", line 384, in __init__
    [get_layer() for _ in range(num_layers)])
  File "/data/Megatron-LM/mpu/transformer.py", line 384, in <listcomp>
    [get_layer() for _ in range(num_layers)])
  File "/data/Megatron-LM/mpu/transformer.py", line 380, in get_layer
    output_layer_init_method=output_layer_init_method)
  File "/data/Megatron-LM/mpu/transformer.py", line 259, in __init__
    self.input_layernorm = LayerNorm(hidden_size, eps=layernorm_epsilon)
  File "/usr/local/lib/python3.6/dist-packages/apex/normalization/fused_layer_norm.py", line 133, in __init__
    fused_layer_norm_cuda = importlib.import_module("fused_layer_norm_cuda")
  File "/usr/lib/python3.6/importlib/__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 994, in _gcd_import
  File "<frozen importlib._bootstrap>", line 971, in _find_and_load
  File "<frozen importlib._bootstrap>", line 955, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 658, in _load_unlocked
  File "<frozen importlib._bootstrap>", line 571, in module_from_spec
  File "<frozen importlib._bootstrap_external>", line 922, in create_module
  File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
ImportError: /usr/local/lib/python3.6/dist-packages/fused_layer_norm_cuda.cpython-36m-x86_64-linux-gnu.so: undefined symbol: _ZN6caffe26detail37_typeMetaDataInstance_preallocated_32E

how can I use it in 3090？