zen_of_python.txt

"Tensors are the foundation, grasp them firmly.
Autograd is the magic, wield it wisely.
Eager is better than lazy, explore and experiment.
Simplicity in API, complexity in models.
Explicit models are better than implicit architectures.
Readability in code, reproducibility in results.
Batch is better than one, parallel is better than serial.
GPU acceleration, the wind beneath your model’s wings.
Be generous with comments, your future self will thank you.
Modularity counts, layer by layer builds the castle.
In the face of ambiguity, refer to the docs.
There should be one – and preferably only one – right way to define a layer.
Although that way may not be obvious unless you’ve read the source.
Now is better than never, but understanding beats rushing.
If the training is hard to debug, revisit your model.
If the results are easy to interpret, you’re onto something.
Experimentation is one honking great idea – let’s do more of those!"