🐛 Bug

The builtin getattr function has a third optional parameter called default, returned if the key fetched does not exist. However, the IR emitter does not support this third parameter and supports only the 2 arguments version:

https://github.com/pytorch/pytorch/blob/master/torch/csrc/jit/frontend/ir_emitter.cpp#L2858-L285

Description

Implement a YOLO model and add it to the DJL model zoo

References

• https://arxiv.org/abs/1506.02640

Issue to track tutorial requests:

• Deep Learning with PyTorch: A 60 Minute Blitz - #69
• Sentence Classification - #79

For simple neuron types like leaky integrate and fire neurons it
is possible to analytically describe the maximal membrane voltage
in response to a spike in terms of the membrane and synapse
time constants. It makes sense to scale the weights according to
that factor in order to normalise the initialisation.

The init module has been deprecated, and the recommend approach for generating initial weights is to use the Template.shape method:

>>> from pennylane.templates import StronglyEntanglingLayers
>>> qml.init.strong_ent_layers_normal(n_layers=3, n_wires=2) # deprecated
>>> np.random.random(StronglyEntanglingLayers.shape(n_layers=3, n_wires=2))  # new approach

We should upd

Okay, so this might not exactly be a "good first issue" - it is a little more advanced, but is still very much accessible to newcomers.

Similar to the mygrad.nnet.max_pool function, I would like there to be a mean-pooling layer. That is, a convolution-style windows is strided over the input, an