def test_deformable_conv():
if not is_cuda_available():
return
conv = M.DeformableConv2d(3, 32, 3)
@trace(symbolic=True, capture_as_const=True)
def fwd(data, offset, mask):
return conv(data, offset, mask)
data = Tensor(np.random.random((1, 3, 32, 32)))
offset = Tensor(np.ones((32, 3 * 3 * 2, 30, 30)).astype("int32") * 5)
mask = Tensor(np.ones((32, 3 * 3, 30, 30)).astype("int32"))
out = fwd(data, offset, mask)
check_pygraph_dump(fwd, [data, offset, mask], [out])
def test_batch_convbias():
if is_cuda_available():
return
@trace(symbolic=True, capture_as_const=True)
def fwd(inp, weight, bias):
return F.quantized.batch_conv_bias_activation(
inp, weight, bias, dtype=dtype.qint8(scale=1.0), nonlinear_mode="relu"
)
inp = Tensor(np.random.random((1, 3, 64, 64)), dtype=dtype.qint8(scale=1.0))
weight = Tensor(np.random.random((1, 32, 3, 3, 3)), dtype=dtype.qint8(scale=1.0))
bias = Tensor(np.random.random((1, 32, 1, 1)), dtype=dtype.qint32(scale=1.0))
result = fwd(inp, weight, bias)
check_pygraph_dump(fwd, [inp, weight, bias], [result])