"""
* Describing Variables: :meth:`describe`
* Aniamting Tensors: :meth:`animate`, :meth:`animate_to_video`
* Plotting Tensors: :meth:`plot_3d_tensor_as_3d_plot`
"""
from types import ModuleType, FunctionType, MethodType
import re
import numpy as np
from .o import save_name, O, Ox
#from .variables import get_convis_attribute, has_convis_attribute, full_path
import inspect
from future.utils import iteritems as _iteritems
import torch
from . import variables
def has_convis_attribute(o,k):
return hasattr(o,k)
def get_convis_attribute(o,k,default=None):
return getattr(o,k,default)
plotting_possible = False
plotting_exceptions = []
do_3d_plot = True
try:
import matplotlib
import matplotlib.pylab as plt
from IPython.core import pylabtools as pt
gui, backend = pt.find_gui_and_backend('inline', 'inline')
from IPython.core.pylabtools import activate_matplotlib
activate_matplotlib(backend)
plotting_possible = True
except Exception as e:
plotting_exceptions.append(e)
pass
doc_urls = [
('convis.filters.retina', 'https://jahuth.github.io/convis/docs_retina.html#'),
('convis.filters.spiking', 'https://jahuth.github.io/convis/filters.html#'),
('convis.filters', 'https://jahuth.github.io/convis/docs_filters.html#'),
('convis.models', 'https://jahuth.github.io/convis/docs_models.html#'),
('convis.retina', 'https://jahuth.github.io/convis/docs_retina.html#'),
('convis.base','https://jahuth.github.io/convis/docs.html#'),
('convis.streams','https://jahuth.github.io/convis/docs_streams.html#'),
('convis.samples','https://jahuth.github.io/convis/docs.html#')
]
[docs]def describe(v,**kwargs):
"""All-purpose function to describe many convis objects
The output adapts itself to being displayed in a
notebook or text console.
"""
return _Descriptor(v,**kwargs)
def help(v,**kwargs):
return _Descriptor({'name': get_convis_attribute(v,'name'),
'doc': get_convis_attribute(v,'doc')},**kwargs)
class _Descriptor(object):
def __init__(self,v,**kwargs):
self.v = v
self.kwargs = kwargs
def __repr__(self):
return repr(describe_dict(self.v))
def __str__(self):
return describe_text(self.v)
def _repr_pretty_(self, p, cycle):
p.text(describe_text(self.v))
def _repr_html_(self):
return describe_html(self.v,wrap_in_html=False,**self.kwargs)
def describe_text(v, indent=' '):
pure_indent = len(indent) * ' '
def repr_if_not_str(s):
if type(s) is str:
return s
return re.sub( '\n+', '\n', repr(s) )
if isinstance(v, variables.Parameter) or isinstance(v, variables.Variable):
lines = [indent+'Variable',indent+'----------']
for k in ['name','doc']:
if hasattr(v,k):
lines.append(indent + ' ' + str(k)+': '+str(getattr(v,k)))
if hasattr(v,'__array__'):
lines.append(indent+' value: ' + repr(v.__array__()))
return '\n'.join(lines)
if hasattr(v,'__iteritems__'):
return (indent + ' keys:' + ', '.join([k for k,vv in v.__iteritems__()]) + '\n' +
'\n'.join([indent[:-1] + '---\\ ' + k + '\n'+describe_text(vv,indent=pure_indent+' | ')+'\n' for k,vv in v.__iteritems__()]))
if type(v) in [list, tuple] or hasattr(v,'__iter__'):
try:
return '\n'.join([describe_text(vv,indent=indent+' ') for vv in v])
except:
# Tensor Variables love to raise TypeErrors when iterated over
pass
d = ''
if isinstance(v, ModuleType):
d += str(getattr(v,'__doc__',""))
for f in dir(v):
vv = getattr(v,f)
d += f + '\n' + '-'*len(f) + '\n'
d += str(getattr(vv,'__doc__',str(type(vv))))
return ('\n'.join([indent + line for line in d.split('\n')])).replace('\n\n','\n')
for k in ['name','simple_name','doc','config_key','optimizable','node','save','init','get','set','auto_name']:
if has_convis_attribute(v,k):
d+= str(k) +':'+ repr_if_not_str(get_convis_attribute(v,k)) + '\n'
try:
d+= 'value:' + repr_if_not_str(v.get_value()) + '\n'
except:
pass
try:
d+= 'got: ' + repr_if_not_str(get_convis_attribute(v,'get',(variables.create_context_O(v)))) + '\n'
except:
pass
return ('\n'.join([indent + line for line in d.split('\n')])).replace('\n\n','\n')
def describe_dict(v):
if type(v) in [list, tuple] or hasattr(v,'__iter__'):
try:
return [describe_dict(vv) for vv in v]
except:
# Tensor Variables love to raise TypeErrors when iterated over
pass
d = {}
if isinstance(v, ModuleType):
d['doc'] = str(getattr(v,'__doc__',""))
for f in dir(v):
vv = getattr(v,f)
d[f] = str(getattr(vv,'__doc__',str(type(vv))))
return d
for k in ['name','simple_name','doc','config_key','optimizable','node','save','init','get','set','auto_name']:
if has_convis_attribute(v,k):
d[k] = get_convis_attribute(v,k)
try:
d['value'] = v.get_value()
except:
pass
try:
d['got'] = get_convis_attribute(v,'get',(variables.create_context_O(v)))
except:
pass
return d
def _plot_to_string():
try:
from StringIO import StringIO
make_bytes = lambda x: x.buf
except ImportError:
from io import BytesIO as StringIO
make_bytes = lambda x: x.getbuffer()
try:
from urllib import quote
except:
from urllib.parse import quote
import base64
import matplotlib.pylab as plt
imgdata = StringIO()
plt.savefig(imgdata)
plt.close()
imgdata.seek(0)
image = base64.encodestring(make_bytes(imgdata))
return str(quote(image))
def _tensor_to_html(t,title='',figsize=(5,4),line_figsize=(5,1.5),line_kwargs={},imshow_kwargs={},preamble=True,**other_kwargs):
"""
This function plots/prints numerical objects of 0,1,2,3 and 5 dimensions such that it can be displayed as html.
"""
kwargs = {'title':title,'figsize':figsize,'line_figsize':line_figsize,'line_kwargs':line_kwargs,'imshow_kwargs':imshow_kwargs}
imshow_kwargs['interpolation'] = imshow_kwargs.get('interpolation','nearest')
if type(t) == int or type(t) == float:
return str(t)
elif type(t) == np.ndarray:
if plotting_possible:
import matplotlib.pylab as plt
if len(t.shape) == 0:
return str(t)
if len(t.shape) == 1:
if t.shape[0] == 1:
if preamble is False:
return str(t[0])
return str(t[0]) + ' (shape 1,)'
else:
plt.figure(figsize=line_figsize)
if title != '':
plt.title(title)
plt.plot(t,**line_kwargs)
plt.tight_layout()
if preamble is False:
return "<img src='data:image/png;base64," + _plot_to_string() + "'>"
return "Numpy array "+str(t.shape)+"<br/><img src='data:image/png;base64," + _plot_to_string() + "'>"
elif len(t.shape) == 2:
if t.shape[0] == 1 and t.shape[1] == 1:
if preamble is False:
return str(t[0])
return str(t[0]) + ' (shape 1,1)'
else:
if np.abs(float(t.shape[0] - t.shape[1]))/(t.shape[0]+t.shape[1]) < 0.143:
# for roughly square 2d objects
plt.figure(figsize=figsize)
if title != '':
plt.title(title)
plt.imshow(t,**imshow_kwargs)
plt.colorbar()
else:
plt.figure(figsize=line_figsize)
# for 2d objects with one long side:
if t.shape[0] > t.shape[1]:
plt.plot(t,**line_kwargs)
else:
plt.plot(t.transpose(),**line_kwargs)
plt.tight_layout()
if preamble is False:
return "<img src='data:image/png;base64," + _plot_to_string() + "'>"
return "Numpy array "+str(t.shape)+"<br/><img src='data:image/png;base64," + _plot_to_string() + "'>"
elif len(t.shape) == 3:
if t.shape[0] == 1 and t.shape[1] == 1 and t.shape[2] == 1:
return str(t[0]) + ' (shape 1,1,1)'
else:
legend = ""
if t.shape[0] == 1:
img = _tensor_to_html(t[0,:,:],preamble=False,**kwargs)
elif t.shape[1] == 1 and t.shape[2] == 1:
img = _tensor_to_html(t[:,0,0],preamble=False,**kwargs)
else:
## Todo: Plotting profiles of other dimensions iff they show interesting changes
#
# Right now the profile is plotted if the tensor is sufficiently small.
# But a better way would be to determine if the profile is "interesting"
# and then sampling a few lines from there.
plt.figure(figsize=figsize)
if title != '':
plt.suptitle(title)
plt.subplot(221)
plt.title('mean over time')
plt.imshow(t.mean(0),**imshow_kwargs)
plt.subplot(222)
plt.title('mean over x')
if t.shape[2] <= 30:
plt.plot(t.mean(0),range(t.shape[1]),'k',alpha=0.2,**line_kwargs)
legend += 'Grey lines show the profile of the x dimension. '
if t.shape[0] <= 100:
plt.plot(t.mean(2).transpose(),range(t.shape[1]),'g',alpha=0.2,**line_kwargs)
plt.plot(t.mean((0,2)),range(t.shape[1]),**line_kwargs)
plt.gca().invert_yaxis()
plt.ylabel('y')
plt.subplot(223)
plt.title('mean over y')
if t.shape[1] <= 30:
plt.plot(t.mean(0).transpose(),'r',alpha=0.2,**line_kwargs)
legend += 'Red lines show the profile of the y dimension. '
if t.shape[0] <= 100:
plt.plot(t.mean(1).transpose(),'g',alpha=0.2,**line_kwargs)
legend += 'Green lines show different time points. '
plt.plot(t.mean((0,1)),**line_kwargs)
plt.xlabel('x')
plt.subplot(224)
plt.title('mean over x and y')
if t.shape[1]+t.shape[2] <= 2*30:
plt.plot(t.reshape((t.shape[0],-1)),color='orange',alpha=0.1,**line_kwargs)
legend += 'Orange lines show the spatial profile over time. '
plt.plot(t.mean((1,2)),**line_kwargs)
plt.xlabel('time')
plt.tight_layout()
img = "<img src='data:image/png;base64," + _plot_to_string() + "'>"
if do_3d_plot and t.shape[0] <= 20:
# TODO: create config for these cutoff values
with OrthographicWrapper():
ax = plot_3d_tensor_as_3d_plot(t)
plt.tight_layout()
img += "<img src='data:image/png;base64," + _plot_to_string() + "'>"
if preamble is False:
return img
return "Numpy array "+str(t.shape)+"<br/>"+img+legend
elif len(t.shape) == 5:
# we assume its actually 3d with extra dimensions
if t.shape[0] == 1 and t.shape[2] == 1:
return "Numpy array "+str(t.shape)+"<br/>"+_tensor_to_html(t[0,:,0,:,:],preamble=False,**kwargs)
else:
if t.shape[0] == 1:
s = "Numpy array "+str(t.shape)+"<br/>"
for i in range(t.shape[2]):
s+= _tensor_to_html(t[0,:,i,:,:],preamble=False,**kwargs) + "<br/>"
return s
elif t.shape[2] == 1:
s = "Numpy array "+str(t.shape)+"<br/>"
for i in range(t.shape[0]):
s+= _tensor_to_html(t[i,:,0,:,:],preamble=False,**kwargs) + "<br/>"
return s
else:
return '5D tensor with too large first or third dimensions!'
return 'Numpy Array (shape '+str(t.shape)+')'
else:
if preamble is False:
return str(t[0])
return str(type(t))+': '+str(t)
on_click_toggle = """onclick='$(this).parent().children(".description_content").toggle();$(this).parent().children(".description_content_replacer").toggle();'"""
var_name_counter = 0
[docs]def animate(ar,skip=10,interval=100):
"""animates a 3d or 5d array in a jupyter notebook
Returns a `matplotlib` animation object.
Parameters
----------
ar (np.array):
3d or 5d array to animate
skip (int):
the animation skips this many timesteps
between two frames.
When generating an html plot or video for long
sequences, this should be set to a higher value
to keep the video short
interval (int):
number of milliseconds between two animation
frames
Examples
--------
To use in a jupyter notebook, use a suitable matplotlib backend::
>>> %matplotlib notebook
>>> import convis
>>> inp = convis.samples.moving_grating(5000)
>>> convis.animate(inp)
<matplotlib.animation.FuncAnimation at 0x7f99d1d88750>
To get a html embedded video, use `convis.variable_describe.animate_to_video`
>>> %matplotlib notebook
>>> import convis
>>> inp = convis.samples.moving_grating(5000)
>>> convis.variable_describe.animate_to_video(inp)
<HTML video embedded in the notebook>
See Also
--------
convis.variable_describe.animate_double_plot
convis.variable_describe.animate_to_video
convis.variable_describe.animate_to_html
"""
import matplotlib.pyplot as plt
from matplotlib.animation import FuncAnimation
import numpy as np
assert len(ar.shape) in [3,5], """The input array has to have 3 or 5 dimensions, not %i!
Input dimensions are: %s"""%(len(ar.shape),str(ar.shape))
if len(ar.shape) == 5:
ar = np.concatenate(np.concatenate(ar,axis=-1),axis=-2)
skipped_array = ar[::skip,:,:]
fig, ax = plt.subplots()
ims = plt.imshow(skipped_array[0],vmin=ar.min(),vmax=ar.max())
plt.colorbar()
plt.title(str(0))
def update(i):
ims.set_array(skipped_array[i])
plt.title(str(i*skip))
return ims
anim = FuncAnimation(fig, update, frames=len(skipped_array), repeat=True, interval=interval)
return anim
[docs]def animate_double_plot(ar,skip=10,interval=200,window_length = 200):
"""animates two plots to show a 3d or 5d array: a spatial and a temporal scrolling line plot
5d arrays will be converted into 3d arrays by concatenating
the batch and channel dimensions in the x and y spatial dimensions.
Parameters
----------
ar (np.array):
3d or 5d array to animate
skip (int):
the animation skips this many timesteps
between two frames.
When generating an html plot or video for long
sequences, this should be set to a higher value
to keep the video short
interval (int):
number of milliseconds between two animation
frames
window_length (int):
length of the window displayed in the
scrolling line plot
See Also
--------
convis.variable_describe.animate
"""
import matplotlib.pyplot as plt
from matplotlib.animation import FuncAnimation
import numpy as np
import convis
assert len(ar.shape) in [3,5], """The input array has to have 3 or 5 dimensions, not %i!
Input dimensions are: %s"""%(len(ar.shape),str(ar.shape))
if len(ar.shape) == 5:
ar = np.concatenate(np.concatenate(ar,axis=-1),axis=-2)
skipped_array = ar[::skip,:,:]
fig, (ax1,ax2) = plt.subplots(1,2)
plt.subplot(ax1)
ims = plt.imshow(skipped_array[0],vmin=ar.min(),vmax=ar.max())
plt_x = np.linspace(0,window_length,window_length)
plt_mean = ax2.plot(ar[:,:,:].mean((1,2)))
plt_examples = ax2.plot(ar[:,int(ar.shape[1]/2),int(ar.shape[2]/2)],alpha=0.5,color='black')
ar_mean = ar[:,:,:].mean((1,2))
ar_min = ar[:,:,:].min((1,2))
ar_max = ar[:,:,:].max((1,2))
plt_fill = ax2.fill_between(np.arange(len(ar_min)),ar_min,ar_max,alpha=0.5)
ar_std_min = ar_mean - ar[:,:,:].std((1,2))
ar_std_max = ar_mean + ar[:,:,:].std((1,2))
plt_pointer = ax2.axvline(0,color='red')
plt_fill = ax2.fill_between(np.arange(len(ar_min)),ar_std_min,ar_std_max,alpha=0.5)
ax2.set_xlim(-0.5*window_length,0.5*window_length)
plt.colorbar()
plt.title(str(0))
plt.tight_layout()
def update(i):
ims.set_array(skipped_array[i])
plt.title(str(i*skip))
ax2.set_xlim(i*skip-0.5*window_length,(i*skip+0.5*window_length))
plt_pointer.set_data([i*skip,i*skip],[-100,100])
return ims,plt_mean
anim = FuncAnimation(fig, update, frames=len(skipped_array), repeat=True, interval=interval)
return anim
[docs]def animate_to_video(ar,skip=10,interval=100,scrolling_plot=False,window_length=200):
"""animates a 3d or 5d array in a jupyter notebook
Returns a Jupyter HTML object containing an embedded video
that can be downloaded.
Parameters
----------
ar (np.array):
3d or 5d array to animate
skip (int):
the animation skips this many timesteps
between two frames.
When generating an html plot or video for long
sequences, this should be set to a higher value
to keep the video short
interval (int):
number of milliseconds between two animation
frames
scrolling_plot (bool):
whether to plot the spatial and temporal plots
or only the spatial animation
window_length (int):
if `scrolling_plot` is `True`, specifies
the length of the time window displayed
Examples
--------
>>> %matplotlib notebook
>>> import convis
>>> inp = convis.samples.moving_grating(5000)
>>> convis.variable_describe.animate_to_video(inp)
<HTML video embedded in the notebook>
See Also
--------
convis.variable_describe.animate
"""
if scrolling_plot:
anim = animate_double_plot(ar,skip=skip,interval=interval,window_length=window_length)
else:
anim = animate(ar,skip=skip,interval=interval)
from IPython.display import HTML
return HTML(anim.to_html5_video())
[docs]def animate_to_html(ar,skip=10,interval=100,scrolling_plot=False,window_length=200):
"""animates a 3d or 5d array in a jupyter notebook
Returns a Jupyter HTML object containing an embedded
javascript animated plot.
Parameters
----------
ar (np.array):
3d or 5d array to animate
skip (int):
the animation skips this many timesteps
between two frames.
When generating an html plot or video for long
sequences, this should be set to a higher value
to keep the video short
interval (int):
number of milliseconds between two animation
frames
scrolling_plot (bool):
whether to plot the spatial and temporal plots
or only the spatial animation
window_length (int):
if `scrolling_plot` is `True`, specifies
the length of the time window displayed
Examples
--------
>>> %matplotlib notebook
>>> import convis
>>> inp = convis.samples.moving_grating(5000)
>>> convis.variable_describe.animate_to_html(inp)
<HTML javascript plot embedded in the notebook>
See Also
--------
convis.variable_describe.animate
"""
if scrolling_plot:
anim = animate_double_plot(ar,skip=skip,interval=interval,window_length=window_length)
else:
anim = animate(ar,skip=skip,interval=interval)
from IPython.display import HTML
return HTML(anim.to_jshtml())
def describe_html(v,wrap_in_html=True,**kwargs):
from IPython.display import HTML
import uuid
try:
import html
except:
import cgi as html # fallback escape function
if isinstance(v, variables.Parameter) or isinstance(v, variables.Variable) or isinstance(v, torch.nn.parameter.Parameter):
d = {}
for k in ['name','_name','simple_name','doc','config_key','optimizable','node','save','init','get','set','variable_type','auto_name']:
if has_convis_attribute(v,k):
d[k] = get_convis_attribute(v,k)
name = d.get('name','') # optional: None handling
if not type(name) is str or name is '':
name= d.get('_name','')
if not type(name) is str or name is '':
name= repr(v)
if type(v) is torch.nn.parameter.Parameter:
name = 'torch.nn.Parameter'
if has_convis_attribute(v,'html_name'):
name+=' '+str(get_convis_attribute(v,'html_name'))
#simple_name = str(d.get('simple_name',''))
s = """<div class='convis_description variable'><b """+on_click_toggle+""">"""+name+"""</b> <small>"""+d.get('variable_type','')+"""</small>"""
# default: show everything, hide on click;
s += "<div class='description_content_replacer' style='border-left: 2px solid #eee; padding-left: 5px; margin-bottom: 10px; display: none;'>(…)</div>"
s += "<div class='description_content' style='border-left: 2px solid #eee; border-top: 2px solid #f8f8f8; padding-left: 5px; margin-bottom: 10px; margin-top: 2px;'>"
if has_convis_attribute(v,'path'):
s += "<small>" + full_path(v) + "</small><br/>"
if has_convis_attribute(v,'doc') and get_convis_attribute(v,'doc') != '':
s += '<p class="doc" style="padding:2px;">'+get_convis_attribute(v,'doc')+'</p>'
if has_convis_attribute(v,'owner'):
s += "<tt style='color: gray;'><small>" + str(v.owner) + "</small></tt><br/>"
for k in ['auto_name','config_key','optimizable','node','save','init','get','set','state_out_state','param_init','state_init','state_in_state','copied_from','config_key','config_default']:
if has_convis_attribute(v,k):
if isinstance(get_convis_attribute(v,k), MethodType):
s+= '<div><b>'+str(k)+'</b>: <tt>method</tt></div>'
elif isinstance(get_convis_attribute(v,k), FunctionType):
s+= '<div><b>'+str(k)+'</b>: <tt>function</tt></div>'
else:
s+= '<div><b>'+str(k)+'</b>: <tt>'+html.escape(str(get_convis_attribute(v,k)))+'</tt></div>'
try:
if hasattr(v,'get_value'):
s+= '<b>value</b>: ' + str(_tensor_to_html(v.get_value(),title=name,**kwargs))
except Exception as e:
s+= '<b>value</b>: ' + str(e)
pass
try:
s+= '<b>got</b>: ' + _tensor_to_html(get_convis_attribute(v,'get')(variables.create_context_O(v)),title=name,**kwargs)
except:
pass
vv = v
if hasattr(v,'detach'):
vv = v.detach()
if hasattr(vv,'__array__'):
s+= '<b>value </b>: ' + _tensor_to_html(vv.__array__())
elif hasattr(vv,'data'):
s+= '<b>value </b>: ' + _tensor_to_html(vv.data.__array__())
s += """</div>"""
s += """</div>"""
if not wrap_in_html:
return s
return HTML(s)
##
# Handeling other datatypes
#
if type(v) == int or type(v) == float:
s = str(v)
if not wrap_in_html:
return s
return HTML(s)
elif hasattr(v,'__array__'):
if hasattr(v,'detach'):
s = _tensor_to_html(v.detach().__array__(),**kwargs)
else:
s = _tensor_to_html(v.__array__(),**kwargs)
if not wrap_in_html:
return s
return HTML(s)
if isinstance(v, torch.nn.Module):
return describe_layer_with_html(v, 4, wrap_in_html)
if isinstance(v, ModuleType):
uid = uuid.uuid4().hex
s = """<div class='convis_description module'><b """+on_click_toggle+""">"""+getattr(v,'__name__','(nameless module)')+"""</b>"""
s += "<div class='description_content_replacer' style='border-left: 2px solid #eee; padding-left: 5px; margin-bottom: 10px; display: none;'>(…)</div>"
s += "<div class='description_content' style='border-left: 2px solid #eee; border-top: 2px solid #f8f8f8; padding-left: 5px; margin-bottom: 10px; margin-top: 2px;'>"
s += '<pre>'+str(getattr(v,'__doc__',"(no doc string found)"))+'</pre>'
for f in dir(v):
if f.startswith('_'):
continue
vv = getattr(v,f)
if isinstance(vv, ModuleType):
s += "<div class='convis_description dict_item'><b>"+str(f)+"</b> (module "+str(getattr(vv,'__name__',''))+")</div>"
continue
s += "<div class='convis_description dict_item'><b id="+uid+save_name(f)+" "+on_click_toggle+" >"+str(f)+"</b> <a style=\"text-decoration: none;\" href='#"+uid+"''>↩</a>"
s += "<div class='description_content_replacer' style='border-left: 0px solid #ddd; padding-left: 5px; display: none;'>(…)</div>"
s += "<div class='description_content' style='border-left: 0px solid #ddd; padding-left: 5px;'>"
s += describe_html(vv,wrap_in_html=False,**kwargs)
s += "</div>"
s += "</div>"
s += """</div>"""
s += """</div>"""
if not wrap_in_html:
return s
return HTML(s)
if isinstance(v, FunctionType):
s = """<div class='convis_description module'><b """+on_click_toggle+""">"""+getattr(v,'__name__','(nameless function)')+"""</b>"""
s += "<div class='description_content_replacer' style='border-left: 2px solid #eee; padding-left: 5px; margin-bottom: 10px; display: none;'>(…)</div>"
s += "<div class='description_content' style='border-left: 2px solid #eee; border-top: 2px solid #f8f8f8; padding-left: 5px; margin-bottom: 10px; margin-top: 2px;'>"
s += '<pre>'+str(getattr(v,'__doc__',"(no doc string found)"))+'</pre>'
s += """</div>"""
s += """</div>"""
if not wrap_in_html:
return s
return HTML(s)
if inspect.isclass(v):
s = """<div class='convis_description module'><b """+on_click_toggle+""">"""+getattr(v,'__name__','(nameless class)')+"""</b>"""
s += "<div class='description_content_replacer' style='border-left: 2px solid #eee; padding-left: 5px; margin-bottom: 10px; display: none;'>(…)</div>"
s += "<div class='description_content' style='border-left: 2px solid #eee; border-top: 2px solid #f8f8f8; padding-left: 5px; margin-bottom: 10px; margin-top: 2px;'>"
s += '<pre>'+str(getattr(v,'__doc__',"(no doc string found)"))+'</pre>'
s += """</div>"""
s += """</div>"""
if not wrap_in_html:
return s
return HTML(s)
if type(v) in [dict] or hasattr(v,'__iteritems__'):
uid = uuid.uuid4().hex
s = "<div class='convis_description list'>"
iteration = list(v.__iteritems__() if hasattr(v,'__iteritems__') else _iteritems(v))
s += "<b id="+uid+" "+on_click_toggle+" >+</b> "
for (k,vv) in iteration:
s += '| <a style="text-decoration: none; font-size: 8pt;" href="#'+uid+save_name(k)+'">'+str(k)+'</a> '
s += "<div class='description_content_replacer' style='border-left: 4px solid #f0f0f0; border-top: 4px solid #f8f8f8; padding-left: 10px; margin-bottom: 10px; display: none;'>(…)</div>"
s += "<div class='description_content' style='border-left: 4px solid #f0f0f0; border-top: 4px solid #f8f8f8; padding-left: 10px; margin-bottom: 10px;'>"
iteration = list(v.__iteritems__() if hasattr(v,'__iteritems__') else _iteritems(v))
path = kwargs.pop('path','')
for (k,vv) in iteration:
s += "<div class='convis_description dict_item'><small>"+path+".</small><b id="+uid+save_name(k)+" "+on_click_toggle+" >"+str(k)+"</b> <a style=\"text-decoration: none;\" href='#"+uid+"''>↩</a>"
s += "<div class='description_content_replacer' style='border-left: 0px solid #ddd; padding-left: 5px; display: none;'>(…)</div>"
s += "<div class='description_content' style='border-left: 0px solid #ddd; padding-left: 5px;'>"
s += describe_html(vv,wrap_in_html=False,path = path + '.' + k, **kwargs)
s += "</div>"
s += "</div>"
s += "</div>"
s += "</div>"
if not wrap_in_html:
return s
return HTML(s)
if type(v) in [list, tuple] or hasattr(v,'__iter__'):
try:
s = "<div class='convis_description list'><b "+on_click_toggle+">List ("+str(len(v))+"):</b>"
s += "<div class='description_content_replacer' style='border-left: 4px solid #f0f0f0; border-top: 4px solid #f8f8f8; padding-left: 10px; margin-bottom: 10px; display: none;'>(…)</div>"
s += "<div class='description_content' style='border-left: 4px solid #f0f0f0; border-top: 4px solid #f8f8f8; padding-left: 10px; margin-bottom: 10px;'>"
s += '\n'.join([describe_html(vv,wrap_in_html=False,**kwargs) for vv in v])
s += "</div>"
s += "</div>"
if not wrap_in_html:
return s
return HTML(s)
except:
# Tensor Variables love to raise TypeErrors when iterated over
pass
##
# Assuming its a annotated variable:
#
d = {}
for k in ['name','simple_name','doc','config_key','optimizable','node','save','init','get','set','variable_type','auto_name']:
if has_convis_attribute(v,k):
d[k] = get_convis_attribute(v,k)
name = d.get('name','') # optional: None handling
if not type(name) is str or name is '':
name= repr(v)
if has_convis_attribute(v,'html_name'):
name+=' '+str(get_convis_attribute(v,'html_name'))
#simple_name = str(d.get('simple_name',''))
s = """<div class='convis_description variable'><b """+on_click_toggle+""">"""+name+"""</b> <small>"""+d.get('variable_type','')+"""</small>"""
# default: show everything, hide on click;
s += "<div class='description_content_replacer' style='border-left: 2px solid #eee; padding-left: 5px; margin-bottom: 10px; display: none;'>(…)</div>"
s += "<div class='description_content' style='border-left: 2px solid #eee; border-top: 2px solid #f8f8f8; padding-left: 5px; margin-bottom: 10px; margin-top: 2px;'>"
if has_convis_attribute(v,'path'):
s += "<small>" + full_path(v) + "</small><br/>"
if has_convis_attribute(v,'doc') and get_convis_attribute(v,'doc') != '':
s += '<p class="doc" style="padding:2px;">'+get_convis_attribute(v,'doc')+'</p>'
if has_convis_attribute(v,'owner'):
s += "<tt style='color: gray;'><small>" + str(v.owner) + "</small></tt><br/>"
for k in ['auto_name','config_key','optimizable','node','save','init','get','set','state_out_state','param_init','state_init','state_in_state','copied_from','config_key','config_default']:
if has_convis_attribute(v,k):
if isinstance(get_convis_attribute(v,k), MethodType):
s+= '<div><b>'+str(k)+'</b>: <tt>method</tt></div>'
elif isinstance(get_convis_attribute(v,k), FunctionType):
s+= '<div><b>'+str(k)+'</b>: <tt>function</tt></div>'
else:
s+= '<div><b>'+str(k)+'</b>: <tt>'+html.escape(str(get_convis_attribute(v,k)))+'</tt></div>'
try:
if hasattr(v,'get_value'):
s+= '<b>value</b>: ' + str(_tensor_to_html(v.get_value(),title=name,**kwargs))
except Exception as e:
s+= '<b>value</b>: ' + str(e)
pass
try:
s+= '<b>got</b>: ' + _tensor_to_html(get_convis_attribute(v,'get')(variables.create_context_O(v)),title=name,**kwargs)
except:
pass
s += """</div>"""
s += """</div>"""
if not wrap_in_html:
return s
return HTML(s)
[docs]class OrthographicWrapper():
"""
This context manager overwrites the persp_transformation function of proj3d
to perform orthographic projections.
Plots that are show()n or save()d in this context will use the projection.
After the context closes, the old projection is restored.
Examples
--------
>>> with convis.variable_describe.OrthographicWrapper():
.. plot_3d_tensor_as_3d_plot(ar) # orthographic projection
>>> plot_3d_tensor_as_3d_plot(ar) # returns to default projection
>>> orth = convis.variable_describe.OrthographicWrapper():
>>> with orth:
.. plot_3d_tensor_as_3d_plot(ar)
>>> plot_3d_tensor_as_3d_plot(ar)
"""
def __init__(self):
pass
def __enter__(self):
from mpl_toolkits.mplot3d import proj3d
def orthogonal_proj(zfront, zback):
a = (zfront+zback)/(zfront-zback)
b = -2*(zfront*zback)/(zfront-zback)
# -0.0001 added for numerical stability as suggested in:
# http://stackoverflow.com/questions/23840756
return np.array([[1,0,0,0],
[0,1,0,0],
[0,0,a,b],
[0,0,-0.0001,zback]])
if not hasattr(proj3d,'old_persp_transformation'):
proj3d.old_persp_transformation = proj3d.persp_transformation
proj3d.persp_transformation = orthogonal_proj
def __exit__(self, eType, eValue, eTrace):
from mpl_toolkits.mplot3d import proj3d
plt.show() # we have to show all the plots before restoring the transformation
if hasattr(proj3d,'old_persp_transformation'):
proj3d.persp_transformation = proj3d.old_persp_transformation
[docs]def plot_3d_tensor_as_3d_plot(ar,ax=None,scale_ar=None,num_levels = 20, contour_cmap='coolwarm_r', contourf_cmap='gray', view=(25, 65)):
"""
Until I come up with a 3d contour plot that shows the contours of a volume,
this function visualizes a sequence of images as contour plots stacked on top of each other.
The sides of the plot show a projection onto the xz and yz planes (at index 0).
ar:
the array to visualize
ax: (if available)
the matplotlib axis (in projection='3d' mode) from eg. calling subplot
if none is provided, the current axis will be converted to 3d projection
scale_ar:
the array that is used for scaling (usefull if comparing arrays or visualizing only a small section)
num_levels:
number of levels of contours
contour_cmap='coolwarm_r':
color map used for line contours
contourf_cmap='gray':
color map used for surface contour
view:
tuple of two floats that give the azimuth and angle of the projection
returns:
axis that was used for plotting
"""
from mpl_toolkits.mplot3d import proj3d
if ax is None:
ax = plt.gca(projection='3d')
if scale_ar is None:
scale_ar = ar
k_range = 0.5*np.max(scale_ar)-np.min(scale_ar)
X, Y = np.meshgrid(np.arange(ar.shape[1]), np.arange(ar.shape[2]))
levels = np.linspace(np.min(scale_ar), np.max(scale_ar), num_levels)
for i,k in enumerate(ar):
alpha = 100.0*(np.std(k)+0.003)
if alpha > 1.0:
alpha = 1.0
if alpha < 0.0:
alpha = 0.0
k_scale = 1.0
ax.contourf(X, Y, i+k_scale*k, cmap=contourf_cmap,
levels=i+k_scale*levels,alpha=0.5*alpha)
ax.contour(X, Y,i-k_scale*k/k_range, cmap=contour_cmap,
levels=i-k_scale*levels/k_range,alpha=1.0,linewidths=3)
side_line_color = [plt.cm.get_cmap(contour_cmap)(1.0-(np.mean(k)-np.min(scale_ar))/k_range)]
cset = ax.contour(X, Y, i*np.ones_like(k), zdir='x', offset=0, cmap=None, colors='k',alpha=0.25)
cset = ax.contour(X, Y, i*np.ones_like(k), zdir='y', offset=0, cmap=None, colors='k',alpha=0.25)
cset = ax.contour(X, Y, i-k_scale*k/k_range, zdir='x', offset=0, cmap=None, colors=side_line_color,alpha=0.5*alpha)
cset = ax.contour(X, Y, i-k_scale*k/k_range, zdir='y', offset=0, cmap=None, colors=side_line_color,alpha=0.5*alpha)
ax.view_init(*view)
ax.set_zlabel('time')
ax.set_xlabel('x')
ax.set_ylabel('y')
ax.set_zticks(range(ar.shape[0]),minor=False)
ax.invert_zaxis()
return ax
def describe_layer_with_html(layer, max_depth = 3,wrap_in_html=False):
from IPython.display import HTML
s = ''
def make_table(title,lst, line_style):
from IPython.display import HTML
lst = list(lst)
sub_string = ''
if len(lst) > 0:
#if title is not None:
# sub_string += '<div style="font-weight: bold;">'+title+'</div>'
for name, val in lst:
sub_string += '<div style="'+line_style+'">'+str(name)+' <i>(<b>'+title+':</b> '+str(val.__class__.__name__)
if hasattr(val,'size'):
if len(val.size()) == 1 and val.size()[0] == 1:
sub_string += ' <i style="font-size: 80%;">scalar value</i>'
else:
sub_string += ' <i style="font-size: 80%;">'+str(' x '.join([str(int(i)) for i in val.size()]))+'</i>'
if hasattr(val,'get'):
v = val.get()
if hasattr(v,'size'):
if type(v.size) is int:
sub_string += ' <i style="font-size: 80%;">'+str(v.size)+'</i>'
elif type(v.size) in [list, tuple]:
sub_string += ' <i style="font-size: 80%;">'+str(' x '.join([str(int(i)) for i in v.size]))+'</i>'
else:
sub_string += ' <i style="font-size: 80%;">'+str(' x '.join([str(int(i)) for i in v.size()]))+'</i>'
else:
sub_string += ' <i style="font-size: 80%;">'+str(v)+'</i>'
sub_string += ')</i>'
if hasattr(val,'doc'):
sub_string += '<div style="font-size: 80%;">'+str(val.doc)+'</div>'
sub_string += '</div>'
return sub_string
def r(model,depth):
sub_string =''
sub_string +='<div class="doc_string">'
if hasattr(model,'__doc__'):
try:
sub_string += '<br />'.join(model.__doc__.strip().split('\n')[:1])+''
except:
pass
sub_string += ' <tiny style="font-size:80%;">(see full doc string with `help()`'
if hasattr(model,'__doc_link__'):
sub_string +=' or browse the documentation for <a href="'+model.__doc_link__+'">'+str(model.__class__.__name__)+'</a>'
else:
cls = str(model.__class__.__module__)+'.'+str(model.__class__.__name__)
if cls.startswith('convis'):
sphinx_path = "https://jahuth.github.io/convis/search.html?q="
for doc_package,doc_url in doc_urls:
if cls.startswith(doc_package):
sphinx_path = doc_url
break
sub_string +=' or browse the documentation for <a href="'+sphinx_path+cls+'" style="color:#222;">'+str(model.__class__.__name__)+'</a>'
sub_string += ')</tiny>'
sub_string += '</div>'
if hasattr(model,'_state') and len(model._state) > 0:
sub_string += make_table('State', [(s,getattr(model,s,None)) for s in model._state], 'margin:1px; padding:2px; background-color: #e3eb91; color: #444; margin-left:10px;')
if hasattr(model,'_named_variables') and len(model._named_variables) > 0:
sub_string += make_table('Variable', model._named_variables.items(), 'margin:1px; padding:2px; background-color: #75bea3; color: #444; margin-left:10px;')
if len(list(model.named_children())) > 0:
#sub_string += '<div style="font-weight: bold;">Module</div>'
for mod_name,mod in list(model.named_children()):
if mod_name is '':
continue
bg = '#eae2e5' if depth%2==0 else '#faf2f5'
sub_string += '<div style="margin:1px; padding:2px; background-color: '+bg+'; color: #444; margin-left:10px;"><div>'
sub_string += '<b style:font-size:120%>'+str(mod_name)+'</b> (<b>Layer:</b> '+str(mod.__class__.__name__)+')</div>'
if depth > 0:
#if isinstance(mod, Layer) or isinstance(mod, torch.nn.Module):
if hasattr(mod, 'named_children'):
sub_string += '<div>'+r(mod, depth-1)+'</div>'
sub_string += '</div>'
return sub_string
s += r(layer,max_depth)
if wrap_in_html:
return HTML(s)
return s