[ADD] Rough draft of README

README.md

@@ -1,8 +1,58 @@
 # `memsave_torch`: Lowering PyTorch's Memory Consumption for Selective Differentiation
 
+This package offers drop-in implementations of PyTorch `nn.Module`s.
+They are as fast as their built-in equivalents, but more memory-efficient whenever you want to compute gradients for a sub-set of parameters (i.e. some have `requires_grad=False`).
+You can convert your neural network by calling our converter.
+
+Take a look at the 'Basic example' how it works.
+
+Currently it supports the following layers:
+
+- `memsave_torch.nn.MemSaveConv2d`
+- ...
+
+Also, each layer has a `.from_nn(layer)` function which allows to convert a `torch.nn.Layer` into its memory-saving equivalent.
+
+## Installation
+
 Install (editable):
 ```bash
 pip install -e .
 ```
 
-[Writeup](memsave_torch/writeup.md)
+## Basic example
+
+```python
+X, y = ...
+
+model = Sequential(...)
+loss_func = MSELoss()
+
+# replace all available layers with mem-saving layers
+model = convert(...)
+
+# same as if you were using the original model
+loss = loss_func(model(X), y)
+```
+
+## Further reading
+
+- Link to documentation
+
+- Link to more examples
+
+- Link to paper/experiments folder
+
+[Writeup](memsave_torch/writeup.md)
+
+## How to cite
+
+If this package has benefited you at some point, consider citing
+
+```bibtex
+
+@article{
+  TODO
+}
+
+```