def test_plot_graphviz_temp(self):
dot = dedent("""
digraph D {
A [shape=diamond]
B [shape=box]
C [shape=circle]
A -> B [style=dashed, color=grey]
A -> C [color="black:invis:black"]
A -> D [penwidth=5, arrowhead=none]
}
""")
import matplotlib.pyplot as plt
temp = get_temp_folder(__file__, "temp_plot_graphviz_temp")
img = os.path.join(temp, "dot.png")
dotf = os.path.join(temp, "dot.dot")
fig, ax = plt.subplots(1, 1)
try:
plot_graphviz(dot, ax=ax, temp_dot=dotf, temp_img=img)
except FileNotFoundError as e:
if "No such file or directory: 'dot'" in str(e):
warnings.warn("graphviz not installed: %r." % e)
return
plt.close('all')
self.assertExists(dotf)
self.assertExists(img)
def plot_onnx(onx, ax=None, dpi=300, temp_dot=None, temp_img=None, show=False):
"""
Plots an ONNX graph into a :epkg:`matplotlib` graph.
:param onx: ONNX object, @see cl OnnxInference
:param ax: existing axes
:param dpi: resolution
:param temp_dot: temporary file,
if None, a file is created and removed
:param temp_img: temporary image,
if None, a file is created and removed
:param show: calls `plt.show()`
:return: axes
"""
# delayed import
from pyquickhelper.helpgen.graphviz_helper import plot_graphviz
if ax is None:
ax = plt.gca() # pragma: no cover
elif isinstance(ax, str) and ax == 'new':
_, ax = plt.subplots(1, 1) # pragma: no cover
if not isinstance(onx, OnnxInference):
oinf = OnnxInference(onx, skip_run=True)
else:
oinf = onx # pragma: no cover
dot = oinf.to_dot()
plot_graphviz(dot, dpi=dpi, ax=ax, temp_dot=temp_dot, temp_img=temp_img)
if show:
plt.show() # pragma: no cover
return ax
def test_plot_graphviz(self):
dot = dedent("""
digraph D {
A [shape=diamond]
B [shape=box]
C [shape=circle]
A -> B [style=dashed, color=grey]
A -> C [color="black:invis:black"]
A -> D [penwidth=5, arrowhead=none]
}
""")
import matplotlib.pyplot as plt
fig, ax = plt.subplots(1, 1)
try:
plot_graphviz(dot, ax=ax)
except FileNotFoundError as e:
if "No such file or directory: 'dot'" in str(e):
warnings.warn("graphviz not installed: %r." % e)
return
plt.close('all')
##########################
# Predictions with onnxruntime.
sess = rt.InferenceSession("pipeline_xgboost.onnx",
providers=['CPUExecutionProvider'])
pred_onx = sess.run(None, {"input": X[:5].astype(numpy.float32)})
print("predict", pred_onx[0])
print("predict_proba", pred_onx[1][:1])
#############################
# Final graph
# +++++++++++
oinf = OnnxInference(model_onnx)
ax = plot_graphviz(oinf.to_dot())
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)
#######################################
# Same example with XGBRegressor
# ++++++++++++++++++++++++++++++
update_registered_converter(XGBRegressor, 'XGBoostXGBRegressor',
calculate_linear_regressor_output_shapes,
convert_xgboost)
data = load_diabetes()
x = data.data
y = data.target
X_train, X_test, y_train, _ = train_test_split(x, y, test_size=0.5)
# We can save the model into ONNX
# format and compute the same predictions in many
# platform using :epkg:`onnxruntime`.
####################################
# Python runtime
# ++++++++++++++
#
# A python runtime can be used as well to compute
# the prediction. It is not meant to be used into
# production (it still relies on python), but it is
# useful to investigate why the conversion went wrong.
# It uses module :epkg:`mlprodict`.
oinf = OnnxInference(onx, runtime="python_compiled")
print(oinf)
##########################################
# It works almost the same way.
pred_pyrt = oinf.run({'X': X_test.astype(numpy.float32)})['variable']
print(diff(pred_skl, pred_pyrt))
#############################
# Final graph
# +++++++++++
ax = plot_graphviz(oinf.to_dot(), dpi=100)
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)
