Adding float mnist_simple example #215
Conversation
Signed-off-by: Omar Shrit <omar@avontech.fr>
|
|
Signed-off-by: Omar Shrit <omar@avontech.fr>
Signed-off-by: Omar Shrit <omar@avontech.fr>
There was a problem hiding this comment.
Oh, I never saw this because I still did not have mailman3 working right for the mlpack-git list. Sorry about that... I would have reviewed this earlier otherwise. This is definitely a nice thing to include, but I wonder if calling the example mnist_cnn_fp32 might be more clear for people scrolling through. Alternately, I wonder if just making another file mnist_cnn_fp32.cpp in the mnist_cnn directory would suffice. I don't think I have an opinion either way.
Fingers crossed that mailman3 successfully delivers this message to the mlpack-git list... 😄
mnist_cnn_float/mnist_cnn.cpp
Outdated
|
|
||
| data::Save("model.bin", "model", model, false); | ||
|
|
||
| cout << "Predicting on test set..." << endl; |
There was a problem hiding this comment.
I looked through the test output and it seems like something fails here after this output:
https://dev.azure.com/mlpack/84320e87-76e3-4b6e-8b6e-3adaf6b36eed/_apis/build/builds/9352/logs/15
2023-09-21T12:51:20.9728229Z Predicting on test set...
2023-09-21T12:51:20.9728383Z Example failed: mnist_cnn_float
But, unfortunately, the CI output isn't more than that. Does the whole example run for you successfully locally?
mnist_cnn_float/mnist_cnn.cpp
Outdated
| @@ -0,0 +1,229 @@ | |||
| /** | |||
| * An example of using Convolutional Neural Network (CNN) for | |||
| * solving Digit Recognizer problem from Kaggle website. | |||
There was a problem hiding this comment.
It's probably worth adding a comment pointing out that this is similar to the other example, but in this one we use fp32 precision. (Same for the mnist_simple example.)
Co-authored-by: Ryan Curtin <ryan@ratml.org>
Co-authored-by: Ryan Curtin <ryan@ratml.org>
Signed-off-by: Omar Shrit <omar@avontech.fr>
|
@rcurtin both of the examples has start training as usual, no problem while loading and training the data. |
Signed-off-by: Omar Shrit <omar@avontech.fr>
|
okay I am getting a segmentation fault during the prediction of the mnist_simple, the reason for this is that I forgot to register the layers for float so the model matrix was empty and therefore armadillo throw the error. |
![mlpack-bot[bot]](https://github.com/avatars/in/23934?s=60&v=4){kind=small}
No description provided.