curiousai / tagger Goto Github PK

Hi, I'm trying to create a new unlabeled dataset and had some questions:
https://github.com/CuriousAI/tagger/blob/master/data/shapes.py#L110
As I see masks are only used to calculate AMI score and doesn't take part in training process, right?
What about the "codes", what is it used for? in shapes it's empty and in Freq20-MNIST it seems related to textures.

I am having trouble running the experiments on the GPU inside of the anaconda environment which is installed using the provided install.sh

When I run THEANO_FLAGS=device=cuda0 python ./runner-shapes50k20x20.py, then I get the error:

ERROR (theano.sandbox.gpuarray): pygpu was configured but could not be imported
Traceback (most recent call last):
  File "/home/ricson/.conda/envs/tagger/lib/python2.7/site-packages/theano/sandbox/gpuarray/__init__.py", line 20, in <module>
    import pygpu
ImportError: No module named pygpu

This would be because pygpu is not specified in the environment.yml file. Furthermore, installing pygpu 0.7.5 via conda would result in the error:

ERROR (theano.sandbox.gpuarray): Could not initialize pygpu, support disabled
Traceback (most recent call last):
  File "/home/ricson/.conda/envs/tagger/lib/python2.7/site-packages/theano/sandbox/gpuarray/__init__.py", line 95, in <module>
    init_dev(config.device)
  File "/home/ricson/.conda/envs/tagger/lib/python2.7/site-packages/theano/sandbox/gpuarray/__init__.py", line 46, in init_dev
    "Make sure Theano and libgpuarray/pygpu "
RuntimeError: ('Wrong major API version for gpuarray:', 2, 'Make sure Theano and libgpuarray/pygpu are in sync.')

because it seems like a specific version of pygpu is needed to support theano 0.8.2, and I don't know exactly which version of pygpu is needed.

I am a student, I recently run your program but there was a mistake, I personally think that my version of the machine environment and you are some different, can you run the version of the environment tell me.
Then I put the results of the operation paste, I think the last line is the key to the problem.

thank you!

Building graphs
Traceback (most recent call last):
File "runner-shapes50k20x20.py", line 25, in
experiment = TaggerExperiment(p)
File "/home/jw/git_workspace/tagger/tagger/tagger_exp.py", line 29, in init
self.tagger = Tagger.create_tagger(self.p)
File "/home/jw/git_workspace/tagger/tagger/tagger.py", line 36, in create_tagger
tagger.apply()
File "/home/jw/git_workspace/tagger/tagger/tagger.py", line 128, in apply
self.clean.update(self.apply_tagger(x_only, False))
File "/home/jw/git_workspace/tagger/tagger/tagger.py", line 303, in apply_tagger
ami_score, ami_score_per_sample = self.mask_accuracy(self.masks_unlabeled, m_hat)
File "/home/jw/git_workspace/tagger/tagger/tagger.py", line 439, in mask_accuracy
ami_score_per_sample = ami_score_op(mask_true[0], mask_est)
File "/home/jw/.conda/envs/tagger/lib/python2.7/site-packages/theano/gof/op.py", line 611, in call
node = self.make_node(*inputs, **kwargs)
File "/home/jw/.conda/envs/tagger/lib/python2.7/site-packages/theano/gof/op.py", line 993, in make_node
(str([inp.type for inp in inputs]), str(self.itypes)))
TypeError: We expected inputs of types '[TensorType(float32, 3D), TensorType(float64, 3D)]' but got types '[TensorType(float32, 3D), TensorType(float32, 3D)]'