def classify_many(self, network: Network, points: torch.tensor, out: torch.tensor):
tus.check_tensors(points=(points, (('batch', None), ('input_dim', network.input_dim)), torch.float32),
out=(out, (('batch', points.shape[0]), ('output_dim', network.output_dim)), torch.float32))
network.eval()
with torch.no_grad():
result = network(points)
out.copy_(result)
def classify_many(self, network: Network, points: torch.tensor,
out: torch.tensor):
network.eval()
with torch.no_grad():
result = network(points, self.recurrent_times, None,
self.input_times)
out.copy_(result)
def forward(self, input: Tensor) -> Tensor:
#input = rotate.forward(input)
w, h = input.shape[-1], input.shape[-2]
im_num = input.detach().numpy()
for i in range(4):
img = im_num[i][0]
for j in range(w):
for k in range(h):
img[j][k] = int(img[j][k]*self.margin)
# pyplot.imshow(im_num[0][0])
# pyplot.show()
with torch.no_grad():
input.copy_(torch.tensor(im_num))
x = self._conv_forward(input, self.weight, self.bias)
return x
