Lesson 3 - Official Topic

Part 1 (2020)
142

If I am having difficulty running one of the .nbs, should I start a new post or post to an existing thread? I previously could run fastai last year on this machine, but it says now it doesn’t have enough memory on the GPU for 01_intro this year. Basically just firefox and the docker are running. nvidia-smi shows basically nothing else running.

Running in LXD docker natively with GPU passthrough to a GTX 1080 with ubuntu 18 lts host.

143

Thanks

144

Yep I am not sure either…

1 Like
145

Use a new thread please. You should adjust your batch size as necessary if you have memory issues.

146

Additional note on CPU vs. GPU for deployment:

You might want to build and deploy a desktop application in an enterprise environment. Basically a big use case believe it or not, especially for my customers. :slight_smile:

Dilbert and friends don’t have GPUs on their PCs.

2 Likes
147

If you do need to deploy directly on a phone, you can use PyTorch Mobile as well.

3 Likes
148

Yes, this is not easy and it’s the drawback with Classification in DL today. Hence my question.

1 Like
149

I don’t think you have to train a model to recognize “not-bearness”. You have to train a model to find features that indicate “bearness” and in the absence of “bearness” it would output “not bear”.

For example I can train a model to recognize my face. I don’t have to give it the face of all other people on earth in order to train it.

3 Likes
150

Maybe to see how not sure the model is for a given image say and setting a threshold? Would that approach work?

151

is Caffe2 another way of deploying PyTorch models like ONNX ?

152

2 posts were merged into an existing topic: Lesson 3 - Non-beginner discussion

153

Look into this maybe?:

3 Likes
154

You are talking about binary classification. Sure works and we all know that. Multi label classification is where issue is.

156

As we will see later on, it won’t work with this model which has to produce numbers that add up to one. You need another kinds of model/loss function for this. Stay tune for the lesson about multi-labels problems :wink:

2 Likes
157

I used that approach to offer a ‘second guess’ classification if the % was below a threshold
https://sportsidentifier.azurewebsites.net/

1 Like
158

Installing voila with these commands:

    !pip install voila
    !jupyter serverextension enable voila --sys-prefix

fails in colab with this error:

        Enabling: voila
        - Writing config: /usr/etc/jupyter
            - Validating...
        Error loading server extension voila
              X is voila importable?

Any known solution?

1 Like
159

I wouldn’t have expected it to work in Colab. Remember: Google own fork of jupyter that’s not compatible with the rest of the world (sadly…)

1 Like
160

Sure this the closest how we get to address this presently.

161

ipywidgets are not supported in native colab. Try this:

162

Same problem about biased algorithms is discussed in this book:Algorithms of Oppression
Book by Safiya Noble.

1 Like