docker-hy/ml-kurkkumopo-training

ml-kurkkumopo-training's Issues

Should error early when folder doesnt exist

ml-kurkkumopo-training/train.py

Lines 45 to 56 in d194975

	def get_images(prefix, url):
	imgs = []
	uri = ""
	try:
	img = imageio.imread(url)
	img = imresize(img, (128, 128))
	identifier = str(uuid.uuid4())
	uri = "imgs/" + prefix + identifier + ".jpg"
	imageio.imwrite(uri, img)
	except:
	return
	return uri

The exception catching is too broad. Any filesystem exception is also hidden, which means that in the exercises if someone has configured the volumes wrongly he will get confusing errors that point inside the train function.

kurkkufront_1  | INFO: Accepting connections at http://localhost:3000
kurkkutrain_1  | Using TensorFlow backend.
kurkkutrain_1  | gathering cucumbers...
kurkkutrain_1  | gathering mopeds...
kurkkutrain_1  | Traceback (most recent call last):
kurkkutrain_1  |   File "train.py", line 84, in <module>
kurkkutrain_1  |     train(X, y)
kurkkutrain_1  |   File "train.py", line 41, in train
kurkkutrain_1  |     model.fit(x=X_train, y=y_train, batch_size=128, epochs=5, verbose=1)
kurkkutrain_1  |   File "/usr/local/lib/python3.6/site-packages/keras/engine/training.py", line 952, in fit
kurkkutrain_1  |     batch_size=batch_size)
kurkkutrain_1  |   File "/usr/local/lib/python3.6/site-packages/keras/engine/training.py", line 751, in _standardize_user_data
kurkkutrain_1  |     exception_prefix='input')
kurkkutrain_1  |   File "/usr/local/lib/python3.6/site-packages/keras/engine/training_utils.py", line 128, in standardize_input_data
kurkkutrain_1  |     'with shape ' + str(data_shape))
kurkkutrain_1  | ValueError: Error when checking input: expected conv2d_1_input to have 4 dimensions, but got array with shape (0, 1)
24_kurkkutrain_1 exited with code 1

After a bit of debugging and removing the try catch I got the following error message. Seems that it could be worth catching at least FileNotFoundError and then raise to halt the program to show an error to user immediately when we know that the training will fail because of missing files/folders. Possibly some other exceptions would be good to not catch too.

kurkkufront_1  | INFO: Accepting connections at http://localhost:3000
kurkkutrain_1  | Using TensorFlow backend.
kurkkutrain_1  | gathering cucumbers...
kurkkutrain_1  | multiprocessing.pool.RemoteTraceback: 
kurkkutrain_1  | """
kurkkutrain_1  | Traceback (most recent call last):
kurkkutrain_1  |   File "/usr/local/lib/python3.6/multiprocessing/pool.py", line 119, in worker
kurkkutrain_1  |     result = (True, func(*args, **kwds))
kurkkutrain_1  |   File "/usr/local/lib/python3.6/multiprocessing/pool.py", line 44, in mapstar
kurkkutrain_1  |     return list(map(*args))
kurkkutrain_1  |   File "train.py", line 53, in get_images
kurkkutrain_1  |     imageio.imwrite(uri, img)
kurkkutrain_1  |   File "/usr/local/lib/python3.6/site-packages/imageio/core/functions.py", line 258, in imwrite
kurkkutrain_1  |     writer = get_writer(uri, format, "i", **kwargs)
kurkkutrain_1  |   File "/usr/local/lib/python3.6/site-packages/imageio/core/functions.py", line 174, in get_writer
kurkkutrain_1  |     request = Request(uri, "w" + mode, **kwargs)
kurkkutrain_1  |   File "/usr/local/lib/python3.6/site-packages/imageio/core/request.py", line 125, in __init__
kurkkutrain_1  |     self._parse_uri(uri)
kurkkutrain_1  |   File "/usr/local/lib/python3.6/site-packages/imageio/core/request.py", line 278, in _parse_uri
kurkkutrain_1  |     raise FileNotFoundError("The directory %r does not exist" % dn)
kurkkutrain_1  | FileNotFoundError: The directory '/src/imgs' does not exist
kurkkutrain_1  | """
kurkkutrain_1  | 
kurkkutrain_1  | The above exception was the direct cause of the following exception:
kurkkutrain_1  | 
kurkkutrain_1  | Traceback (most recent call last):
kurkkutrain_1  |   File "train.py", line 69, in <module>
kurkkutrain_1  |     cucumber_imgs = p.map(func, cucumbers["url"])
kurkkutrain_1  |   File "/usr/local/lib/python3.6/multiprocessing/pool.py", line 288, in map
kurkkutrain_1  |     return self._map_async(func, iterable, mapstar, chunksize).get()
kurkkutrain_1  |   File "/usr/local/lib/python3.6/multiprocessing/pool.py", line 670, in get
kurkkutrain_1  |     raise self._value
kurkkutrain_1  | FileNotFoundError: The directory '/src/imgs' does not exist
24_kurkkutrain_1 exited with code 1