pytorch使用中遇到的问题

1. “RuntimeError: Cannot insert a Tensor that requires grad as a constant. Consider making it a parameter or input, or detaching the gradient”

源码是这样写的：

node_connect_list = [
    self.transition_list[stage][branch][pre_index](pre_node) \
        for pre_index, pre_node in enumerate(node_list)
]

其中self.transition_list被我设置成了list类别。改成ModuleList之后就解决了该问题

RuntimeError: kind_.is_prim() INTERNAL ASSERT FAILED at /opt/conda/conda-bld/pytorch_1573049306803/work/torch/csrc/jit/ir.cpp:922, please report a bug to PyTorch. Only prim ops are allowed to not have a registered operator but onnx::Conv doesn't have one either. We don't know if this op has side effects. (hasSideEffects at /opt/conda/conda-bld/pytorch_1573049306803/work/torch/csrc/jit/ir.cpp:922)

Stack Overflow上有人给出方法，将原来的

from tensorboardX import SummaryWriter

改成

from torch.utils.tensorboard import SummaryWriter

即可