I’m using the code
tfms = tfms_from_stats(stats, sz, aug_tfms=None)
data = ImageClassifierData.from_paths(PATH, tfms=tfms, bs=bs, test_name=‘test’)
learn = ConvLearner.pretrained(arch, data, precompute=True, ps=.5, opt_fn = optim.Adam)
Which has worked in the past. I git pulled a week or so ago. Now getting error:
0%| | 0/1 [00:00<?, ?it/s]Traceback (most recent call last):
File “”, line 3, in
learn = ConvLearner.pretrained(arch, data, precompute=True, ps=.5, opt_fn = optim.Adam)
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/conv_learner.py”, line 184, in pretrained
return cls(data, models, precompute, **kwargs)
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/conv_learner.py”, line 170, in init
if precompute: self.save_fc1()
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/conv_learner.py”, line 242, in save_fc1
if self.data.test_dl: predict_to_bcolz(m, self.data.test_dl, test_act)
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/model.py”, line 17, in predict_to_bcolz
for x,*_ in tqdm(gen):
File “/home/chei/anaconda3/envs/fastai/lib/python3.6/site-packages/tqdm/_tqdm.py”, line 930, in iter
for obj in iterable:
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/dataloader.py”, line 88, in iter
for batch in e.map(self.get_batch, c):
File “/home/chei/anaconda3/envs/fastai/lib/python3.6/concurrent/futures/_base.py”, line 586, in result_iterator
yield fs.pop().result()
File “/home/chei/anaconda3/envs/fastai/lib/python3.6/concurrent/futures/_base.py”, line 425, in result
return self.__get_result()
File “/home/chei/anaconda3/envs/fastai/lib/python3.6/concurrent/futures/_base.py”, line 384, in __get_result
raise self._exception
File “/home/chei/anaconda3/envs/fastai/lib/python3.6/concurrent/futures/thread.py”, line 56, in run
result = self.fn(*self.args, **self.kwargs)
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/dataloader.py”, line 75, in get_batch
res = self.np_collate([self.dataset[i] for i in indices])
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/dataloader.py”, line 75, in
res = self.np_collate([self.dataset[i] for i in indices])
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/dataset.py”, line 168, in getitem
return self.get1item(idx)
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/dataset.py”, line 161, in get1item
x,y = self.get_x(idx),self.get_y(idx)
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/dataset.py”, line 245, in get_x
def get_x(self, i): return open_image(os.path.join(self.path, self.fnames[i]))
File “/home/chei/anaconda3/fastai/courses/dl1/fastai/dataset.py”, line 223, in open_image
raise OSError(‘Is a directory: {}’.format(fn))
OSError: Is a directory: /home/chei/anaconda3/fastai/courses/dl1/299-3site/test/c49-M.pat
My train, valid, and test directories all contain images sorted into directories by class (aka the standard method). Is there any reason this arrangement is not ok for the test set? It used to be. I think.