Closed MichelleShih closed 7 years ago
Hi, can you try the fix_progbar
branch and let me know if it fixes the issue?
7 months with no update. Closing.
error when running python evaluate_camvid.py:
jcc@jcc:~/PycharmProjects/reseg$ python evaluate_camvid.py
Starting training...
Error while formatting 'Cost %(cost)f, DD %(DD)f, UD %(UD)f %(shape)s'
{'DD': None, 'UD': None, 'cost': None, 'shape': None}
Traceback (most recent call last):
File "evaluate_camvid.py", line 241, in <module>
'reload': False
File "evaluate_camvid.py", line 110, in main
reload_=params['reload']
File "/home/jcc/PycharmProjects/reseg/reseg.py", line 834, in train
redirect_stdout=True).start()
File "/home/jcc/anaconda2/lib/python2.7/site-packages/progressbar/bar.py", line 606, in start
self.update(self.min_value, force=True)
File "/home/jcc/anaconda2/lib/python2.7/site-packages/progressbar/bar.py", line 562, in update
StdRedirectMixin.update(self, value=value)
File "/home/jcc/anaconda2/lib/python2.7/site-packages/progressbar/bar.py", line 150, in update
DefaultFdMixin.update(self, value=value)
File "/home/jcc/anaconda2/lib/python2.7/site-packages/progressbar/bar.py", line 54, in update
line = converters.to_unicode('\r' + self._format_line())
File "/home/jcc/anaconda2/lib/python2.7/site-packages/progressbar/bar.py", line 493, in _format_line
widgets = ''.join(self._to_unicode(self._format_widgets()))
File "/home/jcc/anaconda2/lib/python2.7/site-packages/progressbar/bar.py", line 468, in _format_widgets
widget_output = converters.to_unicode(widget(self, data))
File "/home/jcc/PycharmProjects/reseg/utils.py", line 319, in __call__
self.format)
File "/home/jcc/anaconda2/lib/python2.7/site-packages/progressbar/widgets.py", line 73, in __call__
return (format or self.format) % data
TypeError: float argument required, not NoneType
Anyone help? @fvisin @marcociccone version information: Name: progressbar2 Version: 3.18.1 --------update-----------------
if using the fix_progbar branch :
Building model ...
Input shape: (None, 360, 480, 3)
Vgg: After vgg: (None, 256, 90, 120)
ReNet: After 2 rnns 1x1@100 and 2 rnns 1x1@100: (None, 200, 90, 120)
ReNet: After 2 rnns 1x1@100 and 2 rnns 1x1@100: (None, 200, 90, 120)
Transposed conv: 2x2 (str 2x2) @ 50:(None, 50, 180, 240)
Transposed conv: 2x2 (str 2x2) @ 50:(None, 50, 360, 480)
Starting training...
Epoch 0/5000 Up 73 Cost 194624.437500, DD 0.000407, UD 5.495290 (None, 360, 480, 3) /home/jcc/anaconda2/lib/python2.7/site-packages/skimage/util/dtype.py:110: UserWarning: Possible precision loss when converting from float64 to uint8
"%s to %s" % (dtypeobj_in, dtypeobj))
train:367/367 |###########################################| 100% Elapsed Time: 0:07:01
valid:101/101 |###########################################| 100% Elapsed Time: 0:01:51
test:233/233 |############################################| 100% Elapsed Time: 0:04:20
Global Accuracies:
Train 0.67235 Valid 0.56024 Test 0.60904
Mean Class Accuracy - Train 0.24453 Valid 0.24097 Test 0.23874
Mean Class iou - Train 0.17433 Valid 0.16082 Test 0.16225
Saved parameters and options in tmp/camvid_models/model_recseg_camvid.npz in 0.133s
Epoch 1 of 5000 took 1219.473s with overall cost 19466051.297
Epoch 1/5000 Up 106 Cost 235080.406250, DD 0.000447, UD 5.558035 (None, 360, 480, 3)
Is there anything weird in the command lines I showed above?
Hi @brisker,
you commented on 2 different (closed) issues and sent me 2 emails with the same exact message in 2 days. Please count to 10 before posting or sending emails and try to wait for at least a week before asking again the same question. We try our best to give support on this repo, but we are also busy working on other projects.
Back to your issue, your log looks fine to me and the code seems to be running as intended. Why do you think it's weird? If something looks odd to you, please open a new issue and be more specific about the problem.
The fix_progbar
branch fixed the issue with the latest version of progbar, so I am going to merge it in master.
I use the Ubuntu 14.04. The ProgressBar I have setup can be import , but doesn't work. In the file reseg.py, the code
outputs an error
I guess if i set up the wrong ProgressBar? Can you give me a reliable one?