线性回归的简洁实现

https://zh.d2l.ai/chapter_linear-networks/linear-regression-concise.html

第二题一直报错,就是numpy的和ndArray的

loss = gluon.loss.HuberLoss()
for epoch in range(num_epochs):
    for X, y in data_iter:
        with autograd.record():
            l = loss(net(X), y)
        l.backward()
        trainer.step(batch_size)
    l = loss(net(features), labels)
    print(f'epoch {epoch + 1}, loss {l.mean().asnumpy():f}')

TypeError: Operator _npx_fully_connected registered in backend is known as
fully_connected in Python. This is a numpy operator which can only accept MXNet numpy ndarrays, while received a legacy ndarray. Please ensure that you have activated numpy semantics by calling npx.set_np() in your code. If you still see this error with numpy semantics activated, please call as_np_ndarray() upon the legacy ndarray to convert it to an MXNet numpy ndarray, and then feed the converted array to this operator.

转换为ndArray报其他错误,要转换numpy的array.转换为numpy的array,说什么要转换为ndarray,我他妈人都傻了,真死锁

兄弟,问题找到了吗,我版本换到最新的也有这个问题

MXNet 某些损失函数不支持里面 DeepNumPy 的结构了,具体看这个 issue~