Project dependencies may have API risk issues

[PyDeps]

Hi, In neural-style, inappropriate dependency versioning constraints can cause risks.

Below are the dependencies and version constraints that the project is using

numpy
Pillow
scipy
tensorflow

The version constraint == will introduce the risk of dependency conflicts because the scope of dependencies is too strict.
The version constraint No Upper Bound and * will introduce the risk of the missing API Error because the latest version of the dependencies may remove some APIs.

After further analysis, in this project,
The version constraint of dependency Pillow can be changed to ==9.2.0.
The version constraint of dependency Pillow can be changed to >=2.0.0,<=9.1.1.

The above modification suggestions can reduce the dependency conflicts as much as possible,
and introduce the latest version as much as possible without calling Error in the projects.

The invocation of the current project includes all the following methods.

The calling methods from the Pillow

PIL.Image.open
PIL.Image.fromarray

The calling methods from the all methods

tensorflow.compat.v1.Variable.eval
loss_vals.items
tensorflow.compat.v1.ConfigProto
int
imresize
key.loss_arrs.append
len
imsave
map
tensor.get_shape
get_loss_vals.items
vgg.load_net
numpy.mean
numpy.matmul
tensorflow.compat.v1.placeholder
fmt_imsave
iteration_times.append
numpy.uint8.styled_grayscale_rgb.astype.Image.fromarray.convert
build_parser.add_argument
functools.reduce
float
_tensor_size
ax.semilogy
numpy.dstack
epsilon.beta2.beta1.learning_rate.tf.train.AdamOptimizer.minimize.run
main
bias.reshape.reshape
collections.OrderedDict.keys
numpy.random.normal
loss.eval
tensorflow.compat.v1.Graph.device
style_losses.append
math.floor
numpy.transpose
time.time
numpy.array
join
numpy.clip
os.path.dirname
gray2rgb.astype
path.Image.open.np.array.astype
re.match
combined_yuv.Image.fromarray.convert
ax.set_ylabel
numpy.empty
tensorflow.compat.v1.random_normal.astype
hms
imread
tensorflow.compat.v1.Graph
argparse.ArgumentParser
scipy.io.loadmat
matplotlib.use
rgb2gray
list
vgg.net_preloaded
stylize.stylize
matplotlib.pyplot.subplots
tensorflow.compat.v1.nn.bias_add
tensorflow.compat.v1.Session
ax.set_xlabel
numpy.uint8.original_image.astype.Image.fromarray.convert
tensorflow.compat.v1.global_variables_initializer
vgg.unprocess
layer.get_shape
collections.OrderedDict.values
epsilon.beta2.beta1.learning_rate.tf.train.AdamOptimizer.minimize
options.checkpoint_output.options.checkpoint_iterations.count
tensorflow.compat.v1.nn.relu
tensorflow.compat.v1.Variable
build_parser
_conv_layer
numpy.reshape
PIL.Image.fromarray.resize
print_progress
vgg.preprocess
isinstance
tensorflow.compat.v1.matmul
val.eval
numpy.dot
fmt.format
range
numpy.zeros
tensorflow.compat.v1.nn.l2_loss
image.eval.reshape
gray2rgb
sum
format
tensorflow.compat.v1.reshape
collections.OrderedDict.items
tensorflow.compat.v1.transpose
layer.net.eval
tensorflow.compat.v1.nn.conv2d
tensorflow.compat.v1.disable_v2_behavior
tensorflow.compat.v1.Graph.as_default
tensorflow.compat.v1.constant
collections.OrderedDict
PIL.Image.fromarray
build_parser.error
tensorflow.compat.v1.nn.avg_pool
IOError
ax.legend
tensorflow.compat.v1.random_normal
build_parser.parse_args
get_loss_vals
sess.run
ValueError
numpy.clip.astype
_pool_layer
loss_vals.keys
img.Image.fromarray.save
content_losses.append
itr.append
tensorflow.compat.v1.train.AdamOptimizer
print
arr.np.clip.astype
numpy.savetxt
enumerate
tensorflow.compat.v1.nn.max_pool
os.path.isfile
PIL.Image.open
numpy.std
img.np.clip.astype
fig.savefig

@developer
Could please help me check this issue?
May I pull a request to fix it?
Thank you very much.

[anishathalye]

Hi! I'm guessing PyDeps is an academic research project?

Adding the correct (and reasonable) lower bounds and adding the right upper bounds seems reasonable. I don't understand why this bot is suggesting Pillow<=9.1.1 as an upper bound, though. If I were to choose something manually, I'd hope that Pillow devs follow semver and do Pillow<10.