def test_tinygrad():
x = Tensor(x_init, device=self.device)
W = Tensor(W_init, device=self.device)
m = Tensor(m_init, device=self.device)
out = x.dot(W).relu()
out = out.logsoftmax()
out = out.mul(m).add(m).sum()
out.backward()
return out.cpu().data, x.grad.cpu().data, W.grad.cpu().data
def __call__(self, x: Tensor) -> Tensor:
"""
>>> x = Tensor.zeros(2, 10)
>>> layer = Dense(10, 5)
>>> y = layer(x)
>>> y.shape
(2, 5)
"""
return x.dot(self.weight) + self.bias
def test_tinygrad():
x = Tensor(x_init)
W = Tensor(W_init)
m = Tensor(m_init)
out = x.dot(W).relu()
out = out.logsoftmax()
out = out.mul(m).add(m).sum()
out.backward()
return out.data, x.grad, W.grad
def test_tinygrad():
x = Tensor(x_init)
W = Tensor(W_init)
m = Tensor(m_init, requires_grad=False)
out = x.dot(W).relu()
out = out.logsoftmax()
out = out.mul(m).add(m).sum()
out.backward()
return out.cpu().data, x.grad.cpu().data, W.grad.cpu().data
class TinyNet():
def __init__(self):
self.x = Tensor(x_init.copy())
self.W = Tensor(W_init.copy())
self.m = Tensor(m_init.copy())
def forward(self):
out = self.x.dot(self.W).relu()
out = out.logsoftmax()
out = out.mul(self.m).add(self.m).sum()
return out
