EfficientVit export to onnx

[cebain]

when I try to export the trained .pth to onnx, it keeps raising error like below, but I can't find any use of adaptive_avg_pool2d in the model, how can i solve this?
torch.onnx.errors.SymbolicValueError: Unsupported: ONNX export of operator adaptive_avg_pool2d, input size not accessible. Please feel free to request support or submit a pull request on PyTorch GitHub: https://github.com/pytorch/pytorch/issues [Caused by the value 'input.1332 defined in (%input.1332 : Float(*, *, *, *, strides=[65536, 256, 16, 1], requires_grad=1, device=cuda:0) = onnx::Add(%input.1312, %3697), scope: model.effientNet.UnetVitUPer::/efficientViT.efficientvit4.EfficientViT::encoder/torch.nn.modules.container.Sequential::blocks4/efficientViT.efficientvit4.EfficientViTBlock::blocks4.4/efficientViT.efficientvit4.Residual::ffn1 # /home/adminroot/Documents/ccc/pycharmProjects/tinyUPerNET/efficientViT/efficientvit4.py:108:0
)' (type 'Tensor') in the TorchScript graph. The containing node has kind 'onnx::Add'.]