def test_variable():
fname = os.path.join(
this_file_dir,
'runtime_analysis/{}/gradients/variable.txt'.format(DIM))
if os.path.exists(fname):
os.remove(fname)
for lib, call in [
(l, c) for l, c in helpers.calls
if c not in [helpers.tf_graph_call, helpers.mx_graph_call]
]:
time_lib = LIB_DICT[lib]
append_to_file(fname, '{}'.format(lib))
x0 = ivy_gen.array([random.uniform(0, 1) for _ in range(DIM)], f=lib)
ivy_grad.variable(x0, f=lib)
ivy_grad_w_time.variable(x0, f=time_lib)
TIMES_DICT.clear()
for _ in range(100):
log_time(fname, 'tb0')
ivy_grad_w_time.variable(x0, f=time_lib)
log_time(fname, 'tb4', time_at_start=True)
log_time(fname, 'tt0')
ivy_grad.variable(x0, f=lib)
log_time(fname, 'tt1', time_at_start=True)
write_times()
append_to_file(fname, 'end of analysis')
def test_execute_with_gradients():
for lib, call in helpers.calls:
if call is helpers.mx_graph_call:
# mxnet symbolic does not support ivy gradient functions
continue
# func with single return val
func = lambda xs_in: (xs_in[0] * xs_in[0])[0]
xs = [ivy_grad.variable(ivy_gen.array([3.], f=lib))]
y, dydxs = call(ivy_grad.execute_with_gradients, func, xs, f=lib)
assert np.allclose(y, np.array(9.))
if call is helpers.np_call:
# numpy doesn't support autodiff
assert dydxs is None
else:
assert np.allclose(dydxs[0], np.array([6.]))
# func with multi return vals
func = lambda xs_in: ((xs_in[0] * xs_in[0])[0], xs_in[0] * 1.5)
xs = [ivy_grad.variable(ivy_gen.array([3.], f=lib))]
y, dydxs, extra_out = call(ivy_grad.execute_with_gradients,
func,
xs,
f=lib)
assert np.allclose(y, np.array(9.))
assert np.allclose(extra_out, np.array([4.5]))
if call is helpers.np_call:
# numpy doesn't support autodiff
assert dydxs is None
else:
assert np.allclose(dydxs[0], np.array([6.]))
# func with multi weights vals
func = lambda xs_in: (xs_in[0] * xs_in[1])[0]
xs = [
ivy_grad.variable(ivy_gen.array([3.], f=lib)),
ivy_grad.variable(ivy_gen.array([5.], f=lib))
]
y, dydxs = call(ivy_grad.execute_with_gradients, func, xs, f=lib)
assert np.allclose(y, np.array(15.))
if call is helpers.np_call:
# numpy doesn't support autodiff
assert dydxs is None
else:
assert np.allclose(dydxs[0], np.array([5.]))
assert np.allclose(dydxs[1], np.array([3.]))
def test_gradient_descent_update():
for lib, call in helpers.calls:
if call is helpers.mx_graph_call:
# mxnet symbolic does not support ivy gradient functions
continue
ws = [ivy_grad.variable(ivy_gen.array([3.], f=lib))]
dcdws = [ivy_gen.array([6.], f=lib)]
w_new = ivy_gen.array(ivy_grad.gradient_descent_update(ws,
dcdws,
0.1,
f=lib)[0],
f=lib)
assert np.allclose(ivy_gen.to_numpy(w_new), np.array([2.4]))
def test_linspace():
for lib, call in helpers.calls:
if call is helpers.mx_graph_call:
# mxnet symbolic does not support linspace
continue
assert np.allclose(call(ivy_gen.linspace, 1, 10, 100, f=lib),
np.linspace(1, 10, 100),
atol=1e-6)
start = ivy_gen.array([[0., 1., 2.]], f=lib)
stop = ivy_gen.array([[1., 2., 3.]], f=lib)
assert np.allclose(call(ivy_gen.linspace, start, stop, 100, f=lib),
np.linspace(np.array([[0., 1., 2.]]),
np.array([[1., 2., 3.]]),
100,
axis=-1),
atol=1e-6)
start = ivy_gen.array([[[-0.1471, 0.4477, 0.2214]]], f=lib)
stop = ivy_gen.array([[[-0.3048, 0.3308, 0.2721]]], f=lib)
res = np.array([[[[-0.1471, 0.4477, 0.2214], [-0.1786, 0.4243, 0.2316],
[-0.2102, 0.4009, 0.2417], [-0.2417, 0.3776, 0.2518],
[-0.2732, 0.3542, 0.2620], [-0.3048, 0.3308,
0.2721]]]])
assert np.allclose(call(ivy_gen.linspace,
start,
stop,
6,
axis=-2,
f=lib),
res,
atol=1e-4)
if call is helpers.torch_call:
start = ivy_grad.variable(start)
stop = ivy_grad.variable(stop)
res = ivy_grad.variable(res)
assert np.allclose(ivy_gen.linspace(start, stop, 6, axis=-2,
f=lib).detach().numpy(),
res,
atol=1e-4)
def test_adam_update():
for lib, call in helpers.calls:
if call is helpers.mx_graph_call:
# mxnet symbolic does not support ivy gradient functions
continue
ws = Container({'w': ivy_grad.variable(ivy_gen.array([3.], f=lib))})
dcdws = Container({'w': ivy_gen.array([6.], f=lib)})
mw = dcdws
vw = dcdws.map(lambda x, _: x**2)
w_new = ivy_gen.array(ivy_grad.adam_update(ws,
dcdws,
0.1,
mw,
vw,
lib.array(1),
f=lib)[0]['w'],
f=lib)
assert np.allclose(ivy_gen.to_numpy(w_new), np.array([2.96837726]))