Adam Gibson
Adam Gibson
@glenn-boudaer1990 ok looking at this, I"ll have to find the associated underlying TF file and map it to this: https://github.com/eclipse/deeplearning4j/blob/c715aea405980eb043420af86c671032bbb78ec6/deeplearning4j/deeplearning4j-modelimport/src/main/java/org/deeplearning4j/nn/modelimport/keras/layers/TFOpLayer.java#L46 What it does is downloads a TF file. More here:...
@timothyjward @saudet with java 17 LTS forcing module-info I added a semi related PR: https://github.com/eclipse/deeplearning4j/pull/9626 I think I've at least solved part of the problem for modularization. I'm playing with...
@timothyjward thanks for the comments that helps a lot. I'll try to address what I can in the upgrade. Regarding 2 and 3: 2. When splitting things up one problem...
@eix128 these kinds of ops will usually be lower priority for us, if you would like to contribute adding an operation we would be happy to walk you through the...
@marquies do you mind just giving me a direct reproducer + test input for the model? That could be a python script that saves the model or an h5 model...
@marquies generally just a saved model and a sample ndarray is all I would need. The input would be something like: ```python model = keras.models.load_model('path/to/location') input = np.ones(..) model.predict(input) ```...
Not quite... I would need whatever this is: https://github.com/marquies/recognizer2/blob/master/src/main/resources/inference.py#L91
@sandylunar yes we would definitely be interested! Could you submit a pull request to the deeplearning4j model zoo? Steps for how to integrate would also be appreciated. Our model zoo...
@sandylunar @basedrhys JFYI for mdoels from the default deeplearning4j model zoo, we would need to upload them to the normal deepleanring4j azure cdn. I would be happy to accept the...
@basedrhys just following up here, we're getting ready for a 1.0 and I was wondering if we could get this contribution in there. Thanks!