如何设置requires_grad_为false(冻结？)PyTorch惰性层？

mw3dktmi 于 2023-03-02 发布在其他

关注(0)|答案(1)|浏览(197)

PyTorch提供了惰性层，例如torch.nn.LazyLinear。我想冻结网络中的一些层，即确保它们不采用渐变步骤。当我尝试.requires_grad_(False)时，收到错误：
ValueError: Attempted to use an uninitialized parameter in <method 'requires_grad_' of 'torch._C._TensorBase' objects>. This error happens when you are using a惰性模块or explicitly manipulating焊炬.nn.参数.未初始化参数objects. When using LazyModules Call正向with a dummy batch to initialize the parameters before calling torch functions
如何冻结惰性层？

pytorch

来源：https://stackoverflow.com/questions/75550160/how-to-set-requires-grad-to-false-freeze-pytorch-lazy-layers