Chapter 1 error in Machine Learning in Elixir: [warning] passing parameter map to initialization is deprecated

CHSurfer

I just picked up the Machine Learning in Elixir book and am already stuck in Chapter 1.

I am working through the first livebook example. My results are slightly different than in the book

#Explorer.DataFrame<
  Polars[150 x 5]
  sepal_length f64 [5.1, 4.9, 4.7, 4.6, 5.0, ...]


#Explorer.DataFrame<
Polars[150 x 5]
sepal_length float [5.1, 4.9, 4.7, 4.6, 5.0, "..."]

I included the below code, as per instructions

Mix.install([
  {:axon, "~> 0.5"},
  {:nx, "~> 0.5"},
  {:explorer, "~> 0.5"},
  {:kino, "~> 0.8"}
])

but my actual version of axon is 0.7, which might be related…

When i try to evaluate this block:

trained_model_state =
  model
  |> Axon.Loop.trainer(:categorical_cross_entropy, :sgd)
  |> Axon.Loop.metric(:accuracy)
  |> Axon.Loop.run(data_stream, %{}, iterations: 500, epochs: 10)

I get the following warning and error:

20:17:36.449 [warning] passing parameter map to initialization is deprecated, use %Axon.ModelState{} instead
Epoch: 0, Batch: 0, accuracy: 0.0250000 loss: 0.0000000

** (ArgumentError) argument at position 3 is not compatible with compiled function template.

%{i: #Nx.Tensor<
s32

, model_state: #Inspect.Error<
got Protocol.UndefinedError with message:

  """
  protocol Enumerable not implemented for #Nx.Tensor<
    f32[3]
  > of type Nx.Defn.TemplateDiff (a struct). This protocol is implemented for the following type(s): Date.Range, Explorer.Series.Iterator, File.Stream, Function, GenEvent.Stream, HashDict, HashSet, IO.Stream, Kino.Control, Kino.Input, Kino.JS.Live, List, Map, MapSet, Range, Stream, Table.Mapper, Table.Zipper
  """

while inspecting:

  %{
    data: %{
      "dense_0" => %{
        "bias" => #Nx.Tensor<
          f32[3]
        >,
        "kernel" => #Nx.Tensor<
          f32[4][3]
        >
      }
    },
    state: %{},
    __struct__: Axon.ModelState,
    parameters: %{"dense_0" => ["bias", "kernel"]},
    frozen_parameters: %{}
  }

Stacktrace:

(elixir 1.17.1) lib/enum.ex:1: Enumerable.impl_for!/1
(elixir 1.17.1) lib/enum.ex:166: Enumerable.reduce/3
(elixir 1.17.1) lib/enum.ex:4423: Enum.reduce/3
(axon 0.7.0) lib/axon/model_state.ex:359: anonymous fn/2 in Inspect.Axon.ModelState.get_param_info/1
(stdlib 6.0) maps.erl:860: :maps.fold_1/4
(axon 0.7.0) lib/axon/model_state.ex:359: anonymous fn/2 in Inspect.Axon.ModelState.get_param_info/1
(stdlib 6.0) maps.erl:860: :maps.fold_1/4
(axon 0.7.0) lib/axon/model_state.ex:320: Inspect.Axon.ModelState.inspect/2

, optimizer_state: {%{scale: #Nx.Tensor<
f32
>}}, loss_scale_state: %{}, loss:
<<<<< Expected <<<<<
#Nx.Tensor<
f32

==========
#Nx.Tensor<
f64

Argument >>>>>
, y_pred: #Nx.Tensor<
f64[120][3] >,
y_true: #Nx.Tensor<
u8[120][3] >
}

(nx 0.9.1) lib/nx/defn.ex:342: anonymous fn/7 in Nx.Defn.compile_flatten/5
(nx 0.9.1) lib/nx/lazy_container.ex:73: anonymous fn/3 in Nx.LazyContainer.Map.traverse/3
(elixir 1.17.1) lib/enum.ex:1829: Enum."-map_reduce/3-lists^mapfoldl/2-0-"/3
(elixir 1.17.1) lib/enum.ex:1829: Enum."-map_reduce/3-lists^mapfoldl/2-0-"/3
(nx 0.9.1) lib/nx/lazy_container.ex:72: Nx.LazyContainer.Map.traverse/3
(nx 0.9.1) lib/nx/defn.ex:339: Nx.Defn.compile_flatten/5
(nx 0.9.1) lib/nx/defn.ex:331: anonymous fn/4 in Nx.Defn.compile/3
#cell:lze5noxdnhytfee2:5: (file)

If we ignore the warning, I guess the message is telling me that it expected an f32 tensor instead of the f64.

Is this due to my version of Axon or am I missing something else. I have run it both on the recommended version of Elixer/erlang (Elixir 1.14.3 (compiled with Erlang/OTP 25)) as well as Elixir 1.17.1 (compiled with Erlang/OTP 27)

3 comments

/nx #machine-learning-in-elixir

0 585 3

2024-10-16 11:06:14 UTC

Showing Posts 1 to 3

Show Best Posts
Show All (oldest first)
Show All (newest first)

joelpaulkoch

Hi, I’m pretty sure this is because of the Axon version.

You can fix the actual version like this (described here):

Mix.install([
  {:axon, "== 0.5.1"},
  {:nx, "~> 0.5"},
  {:explorer, "~> 0.5"},
  {:kino, "~> 0.8"}
])

Note that you must specify the exact version this way.
Can you try that and report back if it works?

In case you’re wondering, there are some changes in Axon from time to time since it’s pre 1.0 but the basic ideas remain the same, so I think it’s still worth following the examples in version 0.5, you can catch up afterwards.

Post #1