def plot(
self,
path: Union[str, None],
plot_integral: bool,
normalize: bool = True,
title: str = None,
) -> Union[Tuple[plt.Figure, plt.Axes], Tuple[None, None]]:
"""
:param path: the path to save an img version of the chart,
None to display the plot
:param plot_integral: True to plot the calculated loss integrals for each layer,
False to plot the averages
:param normalize: normalize the values to a unit distribution (0 mean, 1 std)
:param title: the title to put on the chart
:return: the created figure and axes if path is None, otherwise (None, None)
"""
names = [
"{} ({})".format(res.name, res.id_) if res.id_ is not None else res.name
for res in self._results
]
values = [
res.sparse_integral if plot_integral else res.sparse_average
for res in self._results
]
if normalize:
mean = numpy.mean(values)
std = numpy.std(values)
values = [(val - mean) / std for val in values]
height = round(len(names) / 4) + 3
fig = plt.figure(figsize=(12, height))
ax = fig.add_subplot(111)
if title is not None:
ax.set_title(title)
ax.invert_yaxis()
frame = pandas.DataFrame(
list(zip(names, values)), columns=["Param", "Sensitivity"]
)
frame.plot.barh(ax=ax, x="Param", y="Sensitivity")
plt.gca().invert_yaxis()
if path is None:
plt.show()
return fig, ax
path = clean_path(path)
create_parent_dirs(path)
plt.savefig(path)
plt.close(fig)
return None, None
def load_json(path: str):
"""
:param path: the path to load a previous analysis from
:return: the KSPerfSensitivityAnalysis instance from the json
"""
path = clean_path(path)
with open(path, "r") as file:
objs = json.load(file)
return PruningPerfSensitivityAnalysis.from_dict(objs)
def save_json(self, path: str):
"""
:param path: the path to save the json file at representing the layer
sensitivities
"""
if not path.endswith(".json"):
path += ".json"
path = clean_path(path)
create_parent_dirs(path)
with open(path, "w") as file:
json.dump(self.dict(), file, indent=2)
def plot(
self,
path: Union[str, None],
title: str = None,
) -> Union[Tuple[plt.Figure, plt.Axes], Tuple[None, None]]:
"""
Plot the recorded sensitivity values
:param path: the path for where to save the plot,
if not supplied will display it
:param title: the title of the plot to apply,
defaults to '{plot_loss_key} LR Sensitivity'
:return: the figure and axes if the figure was displayed; else None, None
"""
fig = plt.figure(figsize=(8, 8))
ax = fig.add_subplot(111)
if title is None:
title = ""
elif title == "__default__":
title = "LR Sensitivity"
ax.set_title(title)
ax.set_xlabel("Learning Rate")
ax.set_ylabel("Avg Loss")
frame = pandas.DataFrame.from_records(
[(lr_res["lr"], lr_res["loss_avg"]) for lr_res in self._results],
columns=["Learning Rate", "Avg Loss"],
)
frame.plot(x="Learning Rate", y="Avg Loss", marker=".", logx=True, ax=ax)
if path is None:
plt.show()
return fig, ax
path = clean_path(path)
create_parent_dirs(path)
plt.savefig(path)
plt.close(fig)
return None, None
def plot(
self,
path: Union[str, None],
title: str = None,
) -> Union[Tuple[plt.Figure, plt.Axes], Tuple[None, None]]:
"""
:param path: the path to save an img version of the chart,
None to display the plot
:param title: the title to put on the chart
:return: the created figure and axes if path is None, otherwise (None, None)
"""
names = ["{} ({})".format(res.name, res.id_) for res in self._results]
values = [res.sparse_comparison() for res in self._results]
height = round(len(names) / 4) + 3
fig = plt.figure(figsize=(12, height))
ax = fig.add_subplot(111)
if title is not None:
ax.set_title(title)
ax.invert_yaxis()
frame = pandas.DataFrame(
list(zip(names, values)), columns=["Param", "Sparse comparison (ms)"]
)
frame.plot.barh(ax=ax, x="Param", y="Sparse comparison (ms)")
plt.gca().invert_yaxis()
if path is None:
plt.show()
return fig, ax
path = clean_path(path)
create_parent_dirs(path)
plt.savefig(path)
plt.close(fig)
return None, None