Christian’s first LiveBook test

• https://www.youtube.com/watch?v=RKvqc-UEe34

How to start

Run this from Windows by calling

docker run -p 8080:8080 --env MIX_INSTALL_DIR=/mixcache -v C:/github/chgeuer/mixcache:/mixcache -v C:/github/chgeuer/livebook:/data livebook/livebook
docker run -p 8080:8080 --env MIX_INSTALL_DIR=/mixcache --env LIVEBOOK_IP=0.0.0.0 --env LIVEBOOK_PASSWORD=Secret123gyJFgny8 --env LIVEBOOK_PORT=8080 -v C:/github/chgeuer/mixcache:/mixcache -v C:/github/chgeuer/livebook:/data livebook/livebook

apt add –no-cache make RUN apk update && apk add –virtual build-dependencies build-base gcc wget git

Install dependencies

Mix.install([
  {:exla, "~> 0.1.0-dev", github: "elixir-nx/nx", sparse: "exla", override: true},
  {:nx, "~> 0.1.0-dev", github: "elixir-nx/nx", sparse: "nx", override: true},
  {:axon, "~> 0.1.0-dev", github: "elixir-nx/axon"}
])

Download training data and unpack

{:ok, { _status, _headers, train_body_gz } } = 
    :httpc.request("https://storage.googleapis.com/cvdf-datasets/mnist/train-images-idx3-ubyte.gz")

<<_::32, n_images::32, n_rows::32, n_cols::32, train_body::binary>> = 
   :zlib.gunzip(train_body_gz)

Quickly visualize the bits

train_tensor =
  train_body
  |> Nx.from_binary({:u, 8})
  |> Nx.reshape({n_images, n_rows, n_cols})
  |> Nx.divide(255)

Nx.to_heatmap(train_tensor)

Download the labels

{:ok, { _status, _headers, labels_body_gz } } = 
    :httpc.request("https://storage.googleapis.com/cvdf-datasets/mnist/train-labels-idx1-ubyte.gz")

<<_::32, n_labels::32, label_body::binary>> = 
   :zlib.gunzip(labels_body_gz)

label_body
|> Nx.from_binary({:u, 8})     # [5,0,4,1,...]
|> Nx.reshape({n_labels, 1})   # [[5],[0],[4],[1],...]
|> Nx.equal(Nx.tensor(Enum.to_list(0..9))) # [0,0,0,0,0,1,0,0,0,0], [1,0,0,0,0,0,0,0,0,0], [0,0,0,0,1,0,0,0,0,0], ...

require Axon

model =
   Axon.input({nil, 28, 28})
   |> Axon.flatten()
   |> Axon.dense(128, activation: :sigmoid)
   |> Axon.dense(10, activation: :softmax)