def __init__(
self,
learn: Learner,
do_remove: bool = True,
hMin=-1,
hMax=1,
nBins=100,
useClasses=False, # if true compute histogram of classes in the last layer
liveChart=True, # show live chart of last layer
modulesId=-1, # array of modules to keep
):
self.hMin = hMin or (-hMax)
self.hMax = hMax
self.nBins = nBins
self.zero_bin = math.floor(-self.nBins * self.hMin /
(self.hMax - self.hMin))
self.liveChart = liveChart
self.allModules = [m for m in flatten_model(learn.model)]
self.useClasses = useClasses
modules = self.allModules
if modulesId:
modules = [self.allModules[i] for i in listify(modulesId)]
self.allModules = modules if modules else self.allModules
self.c = learn.data.c # Number of Calsses
super().__init__(learn, modules, do_remove)
def apply_tfms(self, tfms, duration:int=None, size_factor:tuple=(1,1), do_resolve:bool=True, padding_mode:str='reflection'):
tfms = listify(tfms)
size_tfms = [o for o in tfms if isinstance(o.tfm, TfmCrop)]
if do_resolve:
for tfm in tfms:
tfm.resolve()
x = self.clone()
for tfm in tfms:
if tfm in size_tfms:
crop_target = self._get_duration_crop_target(duration)
x = tfm(x, size=crop_target, padding_mode=padding_mode)
else:
x = tfm(x)
# below is the resizing part, `separate from cropping`
if size_factor:
# read target size from size dictionary, passed to transform method, default to own length (no resize)
_, *orig_size = x.shape
orig_size = tuple(orig_size)
# to multiply element-wise
new_size = tuple(np.array(orig_size)*np.array(size_factor))
if new_size != orig_size: x.resize(new_size)
# if x.config is not None: x.config._sr *= new_size/orig_size
return x
def plotActsHist(self,
cols=10,
toDisplay=None,
hScale=.05,
showEpochs=False,
showLayerInfo=True,
aspectAuto=True,
showImage=True):
histsTensor = self.activations_histogram.stats_hist
hists = [histsTensor[i] for i in range(histsTensor.shape[0])]
if toDisplay:
hists = [hists[i] for i in listify(toDisplay)] # optionally focus
n = len(hists)
cols = cols or 3
cols = min(cols, n)
rows = int(math.ceil(n / cols))
fig = plt.figure(figsize=(20, rows * 4.5))
grid = plt.GridSpec(rows,
cols,
figure=fig,
left=None,
bottom=None,
right=None,
top=None,
wspace=.25,
hspace=.25)
for i, l in enumerate(hists):
img = self.getHistImg(l, self.useClasses)
dead = self.getMin(l, self.useClasses, self.zero_bin)
cr = math.floor(i / cols)
cc = i % cols
main_ax = fig.add_subplot(grid[cr, cc])
if showImage: main_ax.imshow(img)
layerId = listify(toDisplay)[i] if toDisplay else i
m = self.allModules[layerId]
outShapeText = f' (out: {list(self.shape_out[m])})' if (
m in self.shape_out) else ''
title = f'L:{layerId}' + '\n' + splitAtFirstParenthesis(
str(m), False, outShapeText)
main_ax.set_title(title, fontsize=8, weight='bold')
imgH = img.shape[0]
main_ax.set_yticks([])
main_ax.set_ylabel(str(self.hMin) + " : " + str(self.hMax))
if aspectAuto: main_ax.set_aspect('auto')
imgW = img.shape[1]
imgH = img.shape[0]
ratioH = -self.hMin / (self.hMax - self.hMin)
zeroPosH = imgH * ratioH
main_ax.plot(dead * l.shape[1], 'r', linewidth=2) # X Axis
main_ax.plot([0, imgW], [zeroPosH, zeroPosH], 'black') # X Axis
if (showEpochs):
start = 0
nEpochs = len(self.activations_histogram.stats_epoch)
for i, hh in enumerate(self.activations_histogram.stats_epoch):
if (i < (nEpochs - 1)):
main_ax.plot([hh, hh], [0, imgH], color=[0, 0, 1])
end = hh # rolling
domain = l[start:end]
domain_mean = domain.mean(-1) # mean on classes
if self.useClasses:
self.plotPerc(main_ax,
domain,
hScale,
1,
start,
colorById=True,
addLabel=(0 == i)) #plot all
main_ax.legend(loc='upper left')
else:
self.plotPerc(main_ax, domain_mean, hScale, .5, start)
self.plotPerc(main_ax,
domain_mean,
hScale,
1,
start,
linewidth=1.5)
start = hh
main_ax.set_xlim([0, imgW])
main_ax.set_ylim([0, imgH])
plt.show()
if showLayerInfo:
for i, l in enumerate(self.allModules):
print('{:2} {}'.format(i, l))