Backto PyTorch Index
本来想自己总结一下,发现一篇好文已经总结的非常好了:Some important Pytorch tasks - A concise summary from a vision researcher: 设计自定义模型中,freeze / update/ delete / add
some layers, 和自定义 loss_function
等,非常直接明了的实战经验。掌握之后,就像作者原文里说的:
Really, once you have done this, you can do anything with model architectures in PyTorch.