docs: add installation details

.github/workflows/Documentation.yml

@@ -16,6 +16,7 @@ on:
 
 # Sets permissions of the GITHUB_TOKEN to allow deployment to GitHub Pages
 permissions:
+ actions: write
  contents: write
  pages: write
  id-token: write

docs/Project.toml

@@ -3,3 +3,4 @@ Boltz = "4544d5e4-abc5-4dea-817f-29e4c205d9c8"
 Documenter = "e30172f5-a6a5-5a46-863b-614d45cd2de4"
 DocumenterVitepress = "4710194d-e776-4893-9690-8d956a29c365"
 Lux = "b2108857-7c20-44ae-9111-449ecde12c47"
+Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"

docs/src/index.md

@@ -44,3 +44,51 @@ features:
  link: https://sciml.ai/
 ---
 ```
+
+## How to Install Boltz.jl?
+
+Its easy to install Boltz.jl. Since Boltz.jl is registered in the Julia General registry,
+you can simply run the following command in the Julia REPL:
+
+```julia
+julia> using Pkg
+julia> Pkg.add("Boltz")
+```
+
+If you want to use the latest unreleased version of Boltz.jl, you can run the following
+command: (in most cases the released version will be same as the version on github)
+
+```julia
+julia> using Pkg
+julia> Pkg.add(url="https://github.com/LuxDL/Boltz.jl")
+```
+
+## Want GPU Support?
+
+Install the following package(s):
+
+:::code-group
+
+```julia [NVIDIA GPUs]
+using Pkg
+Pkg.add("LuxCUDA")
+# or
+Pkg.add(["CUDA", "cuDNN"])
+```
+
+```julia [AMD ROCm GPUs]
+using Pkg
+Pkg.add("AMDGPU")
+```
+
+```julia [Metal M-Series GPUs]
+using Pkg
+Pkg.add("Metal")
+```
+
+```julia [Intel GPUs]
+using Pkg
+Pkg.add("oneAPI")
+```
+
+:::

docs/src/tutorials/getting_started.md

@@ -0,0 +1,56 @@
+# Getting Started
+
+!!! tip "Prerequisites"
+
+ Here we assume that you are familiar with [`Lux.jl`](https://lux.csail.mit.edu/stable/).
+ If not please take a look at the
+ [Lux.jl tutoials](https://lux.csail.mit.edu/stable/tutorials/).
+
+`Boltz.jl` is just like `Lux.jl` but comes with more "batteries included". Let's start by
+defining an MLP model.
+
+```@example getting_started
+using Lux, Boltz, Random
+```
+
+## Multi-Layer Perceptron
+
+If we were to do this in `Lux.jl` we would write the following:
+
+```@example getting_started
+model = Chain(
+ Dense(784, 256, relu),
+ Dense(256, 10)
+)
+```
+
+But in `Boltz.jl` we can do this:
+
+```@example getting_started
+model = Layers.MLP(784, (256, 10), relu)
+```
+
+The `MLP` function is just a convenience wrapper around `Lux.Chain` that constructs a
+multi-layer perceptron with the given number of layers and activation function.
+
+## How about VGG?
+
+!!! warning "Returned Values"
+
+ The returned value from `Vision` module functions are a 3 tuple of (model, ps, st).
+ The `ps` and `st` are the parameters and states of the model respectively.
+
+Let's take a look at the `Vision` module. We can construct a VGG model with the
+following code:
+
+```@example getting_started
+model, _, _ = Vision.VGG(13)
+model
+```
+
+We can also load pretrained ImageNet weights using
+
+```@example getting_started
+model, _, _ = Vision.VGG(13; pretrained=true)
+model
+```