Lesson 1 discussion

martin · December 22, 2016, 11:10pm

Hi, @rachel, may I install Python 3 to study this course, or Python 2 only? Thank you.

I am just on the journey to this course.

oceanos · December 22, 2016, 11:27pm

I,ve commented the import cv2, upgraded to keras 1.2.0 but still stucked in the error “cannot import name ReduceLROOnPlateau” in the “import utils” cell in the notebook.

What should I do?

sarwojowo · December 23, 2016, 8:17am

hi all

i got error when pip install awscli, error like picture below somebody can help me ?

Steve · December 23, 2016, 6:07pm

Rui, I’m having the same issue – same scenario. I’m currently trying to run this against a t2.large instance to save a couple $$ – beginning to think that’s a fool’s errand. i.e. I need to try running it on the p2 instance as @jeremy did in the video. What instance type are you running? Thx!

dschwant · December 23, 2016, 7:02pm

Azure Setup, need more instructions or help…

I know this is brand new, so I am ready to be one of the first to get through this approach,

Here were my steps to get an Azure VM up:

Add new VM

East US
HDD
NC6

Putty to the SSH address given
sh install-gpu-azure.sh

Kind of hung at the point prompted for a password. Not sure if it wants me to set one for something or ?

jupyter doesn’t run yet.

Any help out there on this yet?

jeremy · December 23, 2016, 7:03pm

Yes, it is asking you to set your password, protect your notebook server. Pick whatever you like.

Steve · December 23, 2016, 7:18pm

Quick update: did a graceful showdown of the juypter notebook and the t2 instance. Once I brought it all back up, I no longer had the “cannot import name ReduceLROOnPlateau” error. While it takes a little longer to run on the t2, I was classifying dogs and cats in no time!

jeremy · December 23, 2016, 7:24pm

You can remove this import as well. I will update the files now and push them to github — thanks for letting us know about the issue.

dschwant · December 24, 2016, 4:12am

Jeremy, first let me say thank you very much for all of the hard work you all have put into this course. I know the extra effort and sacrifice you are making and I personlly don’t take it for granted.

So, I assumed that is what it wanted and did try to set a password, but jupyter doesn’t seem to have installed correctly or I am missing something. I will whack this VM and start with a fresh one. I know that since I am trying to do this on an Azure VM then I am a little more on my own as all of the in depth instructions are for AWS. After I try it again I can be more detailed with what I am seeing if it doesn’t work. I am also more than willing to help out with better Azure instructions if I can get this going and someone has already paved the way. Thank you again!!!

complancoder · December 24, 2016, 6:37am

@dschwant I have a put up a post in Medium for helping people set up in Azure . You might wanna look at it .

Can this link be added to the wiki ?

Daniel · December 24, 2016, 9:36am

Thanks Jeremy, it is helpful solution to the import problem.

my question is if I ever need to update or install packages, I will have to do it under ubuntu@ip-10-0-0-8:~$ rather than in my local mac directory. Is it right?

Daniel · December 24, 2016, 1:49pm

Hi Jeremy,

I am running Lesson 1 notebook on mac OS X Sierra version 10.12.1 using t2. Everything runs fine until vgg = Vgg16(), I got the following error:

MemoryError Traceback (most recent call last)
in ()
----> 1 vgg = Vgg16()

/home/ubuntu/nbs/vgg16.pyc in init(self, size, include_top)
31 def init(self, size=(224,224), include_top=True):
32 self.FILE_PATH = ‘http://www.platform.ai/models/’
—> 33 self.create(size, include_top)
34 self.get_classes()
35

/home/ubuntu/nbs/vgg16.pyc in create(self, size, include_top)
84
85 model.add(Flatten())
—> 86 self.FCBlock()
87 self.FCBlock()
88 model.add(Dense(1000, activation=‘softmax’))

/home/ubuntu/nbs/vgg16.pyc in FCBlock(self)
60 def FCBlock(self):
61 model = self.model
—> 62 model.add(Dense(4096, activation=‘relu’))
63 model.add(BatchNormalization())
64 model.add(Dropout(0.5))

/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/models.pyc in add(self, layer)
306 output_shapes=[self.outputs[0]._keras_shape])
307 else:
–> 308 output_tensor = layer(self.outputs[0])
309 if type(output_tensor) is list:
310 raise Exception('All layers in a Sequential model ’

/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/engine/topology.pyc in call(self, x, mask)
485 ‘layer.build(batch_input_shape)’)
486 if len(input_shapes) == 1:
–> 487 self.build(input_shapes[0])
488 else:
489 self.build(input_shapes)

/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/layers/core.pyc in build(self, input_shape)
693
694 self.W = self.init((input_dim, self.output_dim),
–> 695 name=’{}_W’.format(self.name))
696 if self.bias:
697 self.b = K.zeros((self.output_dim,),

/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/initializations.pyc in glorot_uniform(shape, name, dim_ordering)
57 fan_in, fan_out = get_fans(shape, dim_ordering=dim_ordering)
58 s = np.sqrt(6. / (fan_in + fan_out))
—> 59 return uniform(shape, s, name=name)
60
61

/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/initializations.pyc in uniform(shape, scale, name)
30
31 def uniform(shape, scale=0.05, name=None):
—> 32 return K.random_uniform_variable(shape, -scale, scale, name=name)
33
34

/home/ubuntu/anaconda2/lib/python2.7/site-packages/keras/backend/theano_backend.pyc in random_uniform_variable(shape, low, high, dtype, name)
138
139 def random_uniform_variable(shape, low, high, dtype=_FLOATX, name=None):
–> 140 return variable(np.random.uniform(low=low, high=high, size=shape),
141 dtype=dtype, name=name)
142

mtrand.pyx in mtrand.RandomState.uniform (numpy/random/mtrand/mtrand.c:17350)()

mtrand.pyx in mtrand.cont2_array_sc (numpy/random/mtrand/mtrand.c:3092)()

MemoryError:

I have no idea how to get it fixed. Could you have a look?

Thanks

jeremy · December 24, 2016, 5:59pm

Right!

jeremy · December 24, 2016, 6:00pm

This is great! Can we just add you as a wiki editor so you can put the content directly on the wiki?

jeremy · December 24, 2016, 7:15pm

What size t2 are you running it on? Make sure it’s a t2 large. You may need to restart your kernel, and ensure nothing else is running.

alexandrecc · December 24, 2016, 8:25pm

I received today from AWS my p2 limit upgrade after waiting 36h from the initial request. I commented the import cv2 line in utils.py. The ¨pip install git+git://github.com/fchollet/keras.git --upgrade¨, as suggested by ravivijay (thanks btw), solved the “cannot import name ReduceLROnPlateau” error. My lesson 1 is running on p2.xlarge. I’ll try the kaggle homework in the next few days.

Daniel · December 25, 2016, 12:04am

Thanks Jeremy, and Merry Christmas!

I just figured out that he reason why bash setup-t2.sh didn’t get me t2.large is inside the script, it is t2.micro being asked for. So, I changed it to t2.large, then I got t2.large. aws-alias.sh should change t2.micro to t2.large too.

Daniel · December 25, 2016, 12:08am

Hi Alexandre,
How to you check whether you got updated? from the case history webpage where we request for instance?

Thanks

complancoder · December 25, 2016, 11:12am

Sure I can do that!

dacliem · December 25, 2016, 12:41pm

Hi Tom,

Many thanks for your solutions. It works.