Hey folks,
I’m quite new to the ML world, as well as to Nx and Axon (although not necessarily to Elixir). I’ve been trying to follow some articles, as well as Sean Moriarty’s beta book, but I often get stuck with no real idea of what’s happening, so it’d be great if someone could help.
While running the code from How to Create Neural Network in Elixir Using Nx and Axon | Curiosum (with some adaptations since it’s a 1-year-old article), I get to the following error when trying to run Axon.Loop.run:
Reshape operation has mismatched element counts: from=0 (f32[32,0,8,64]) to=16384 (f32[32,512])
For reference, this is what I have so far:
{images, labels} = Scidata.CIFAR10.download()
{img_data, img_type, img_shape} = images
{label_data, label_type, label_shape} = labels
batch_size = 32
images =
img_data
|> Nx.from_binary(img_type)
|> Nx.reshape(img_shape)
|> Nx.divide(255.0)
labels =
label_data
|> Nx.from_binary(label_type)
|> Nx.reshape(label_shape)
|> Nx.new_axis(-1)
|> Nx.equal(Nx.iota({1, 10}))
(...)
train_data =
train_images
|> Nx.to_batched(batch_size)
|> Stream.zip(Nx.to_batched(train_labels, batch_size))
(...)
model =
Axon.input("images", shape: {nil, 3, 32, 32})
|> Axon.conv(32, kernel_size: {3, 3}, activation: :relu, padding: :same)
|> Axon.max_pool(kernel_size: {2, 2}, strides: [2, 2])
|> Axon.conv(64, kernel_size: {3, 3}, activation: :relu, padding: :same)
|> Axon.max_pool(kernel_size: {2, 2}, strides: [2, 2])
|> Axon.flatten()
|> Axon.dense(64, activation: :relu)
|> Axon.dense(10, activation: :softmax)
trained_model_state =
model
|> Axon.Loop.trainer(:categorical_cross_entropy, :adam)
|> Axon.Loop.metric(:accuracy)
|> Axon.Loop.run(train_data, %{}, epochs: 10, compiler: EXLA)
