Ben Morcos
Ben Morcos
Some operators create extra arrays, which can cause memory problems with larger models. In some cases this is avoidable. For example, the [ReLu step math](https://github.com/nengo/nengo/blob/master/nengo/neurons.py#L250) can be modified from it's...
When implementing a _learn every n-th timestep_ feature from #387 it was decided that the `delta` signal should be scaled to help keep learning at a similar rate to the...
If the `LMUCell` is wrapped in another layer (e.g. `RNN`) then it cannot be serialized since `LMUCell` is a custom object unknown to Keras. For example: ```python # Build an...
NengoDL expects a Node with no input connection to receive an input during training/inference even if the node is generating an internal constant. There is a workaround using a lambda...
Right now it's possible to use `tf.keras.callbacks.ModelCheckpoint` to save (keras) model weights during training under NengoDL, and then use `sim.keras_model.load_weights(...)` afterwards. However it might be nice to have our own...
The [net.ipynb](https://github.com/nengo/nengo-gui/blob/6c0f36611feee37d567042676907614a40be4146/examples/net.ipynb) example still uses the deprecated `IPythonViz`, This should probably be updated to the new `InlineGUI` Just noticed this when @ikajic posted the example in the summer school.
Passing in `-d` still prompts to include the `-r` flag despite the target directory being empty. For example: ```bash $ mkdir empty $ rm -d empty Directories ['empty'] included but...
@tensorflow/micro **System information** - Host OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 16.04 - TensorFlow installed from (source or binary): `pip install tensorflow` - Tensorflow version (commit...