def test_estimator_sort_traces_with_available_outputs_have_no_dependency(
self):
trace1 = Trace(inputs="x", outputs="y")
trace2 = Trace(inputs="x", outputs="z")
trace3 = Trace(inputs="x", outputs="x")
with self.subTest("[trace1, trace2]"):
sorted_traces = fe.trace.sort_traces([trace1, trace2],
available_outputs="x")
self.assertEqual(sorted_traces, [trace1, trace2])
with self.subTest("[trace2, trace1]"):
sorted_traces = fe.trace.sort_traces([trace2, trace1],
available_outputs="x")
self.assertEqual(sorted_traces, [trace2, trace1])
with self.subTest("[trace1, trace3]"):
sorted_traces = fe.trace.sort_traces([trace1, trace3],
available_outputs="x")
self.assertEqual(sorted_traces, [trace1, trace3])
with self.subTest("[trace3, trace1]"):
sorted_traces = fe.trace.sort_traces([trace3, trace1],
available_outputs="x")
self.assertEqual(sorted_traces, [trace3, trace1])
def test_estimator_sort_traces_without_available_outputs_have_dependency(
self):
trace1 = Trace(inputs="x", outputs="y")
trace2 = Trace(inputs="y", outputs="z")
trace3 = Trace(inputs="z", outputs="w")
sorted_traces = fe.trace.sort_traces([trace3, trace2, trace1])
self.assertEqual(sorted_traces, [trace1, trace2, trace3])
def test_estimator_sort_traces_cycle_dependency(self):
trace1 = Trace(inputs="x", outputs="y")
trace2 = Trace(inputs="y", outputs="z")
trace3 = Trace(inputs="z", outputs="x")
trace4 = Trace(inputs="x", outputs="a")
with self.subTest("available_outputs='x'"):
sorted_traces = fe.trace.sort_traces(
[trace1, trace2, trace3, trace4], available_outputs="y")
self.assertEqual(sorted_traces, [trace2, trace3, trace4, trace1])
with self.subTest("available_output=None"):
with self.assertRaises(AssertionError):
sorted_traces = fe.trace.sort_traces(
[trace1, trace2, trace3, trace4])
def test_estimator_warmup_tf_dataset_torch_model_smoke(self):
loader = get_sample_tf_dataset(expand_axis=1)
pipeline = fe.Pipeline(train_data=loader) # "x", "y"
model = fe.build(model_fn=LeNetTorch, optimizer_fn="adam")
network = fe.Network(ops=[ModelOp(model=model, inputs="x", outputs="y_pred")])
est = fe.Estimator(pipeline=pipeline, network=network, epochs=1, traces=[Trace(inputs="y_pred")])
est._prepare_traces(run_modes={"train", "eval"})
est._warmup()
self.assertTrue(True)
def test_estimator_warmup_trace_missing_key(self):
loader = get_sample_tf_dataset()
pipeline = fe.Pipeline(train_data=loader) # "x", "y"
model = fe.build(model_fn=LeNetTf, optimizer_fn="adam")
network = fe.Network(
ops=[ModelOp(model=model, inputs="x", outputs="y_pred")])
est = fe.Estimator(pipeline=pipeline,
network=network,
epochs=1,
traces=[Trace(inputs="z")]) # miss key "z"
est._prepare_traces(run_modes={"train", "eval"})
with self.assertRaises(AssertionError):
est._warmup(warmup=True)
def test_estimator_sort_traces_with_available_outputs_no_input_match(self):
trace1 = Trace(inputs="y", outputs="z")
with self.assertRaises(AssertionError):
sorted_traces = fe.trace.sort_traces([trace1],
available_outputs="x")
