Fix typos and add codespell GitHub action (#528)

.github/workflows/test.yaml

@@ -11,6 +11,13 @@ env:
   XLA_CACHE_DIR: ${{ github.workspace }}/cache/xla
   LIBTORCH_DIR: ${{ github.workspace }}/cache/torch
 jobs:
+  codespell:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: codespell-project/actions-codespell@v2
+        with:
+          skip: deps
+          ignore_words_list: whn,ehr
   main:
     runs-on: ubuntu-latest
     steps:

guides/guides.md

@@ -20,7 +20,7 @@ Axon is a library for creating and training neural networks in Elixir. The Axon
 
 * [Your first training loop](training_and_evaluation/your_first_training_loop.livemd)
 * [Instrumenting loops with metrics](training_and_evaluation/instrumenting_loops_with_metrics.livemd)
-* [Your first evalutaion loop](training_and_evaluation/your_first_evaluation_loop.livemd)
+* [Your first evaluation loop](training_and_evaluation/your_first_evaluation_loop.livemd)
 * [Using loop event handlers](training_and_evaluation/using_loop_event_handlers.livemd)
 * [Custom models, loss functions, and optimizers](training_and_evaluation/custom_models_loss_optimizers.livemd)
 * [Writing custom metrics](training_and_evaluation/writing_custom_metrics.livemd)

lib/axon/loop.ex

@@ -935,7 +935,7 @@ defmodule Axon.Loop do
       only: N # Trigger on `N` event
 
   **Warning: If you modify the step state in an event handler, it will trigger
-  potentially excessive recompilation and result in significant additinal overhead
+  potentially excessive recompilation and result in significant additional overhead
   during loop execution.**
   """
   def handle_event(%Loop{handlers: handle_fns} = loop, event, handler, filter \\ :always) do

lib/axon/losses.ex

@@ -221,7 +221,7 @@ defmodule Axon.Losses do
 
   $$l_i = -\sum_i^C \hat{y_i} \cdot \log(y_i)$$
 
-  Categorical cross-entropy is typically used for multi-class classifcation problems.
+  Categorical cross-entropy is typically used for multi-class classification problems.
   By default, it expects `y_pred` to encode a probability distribution along the last
   axis. You can specify `from_logits: true` to indicate `y_pred` is a logits tensor.
 
@@ -962,7 +962,7 @@ defmodule Axon.Losses do
   @doc """
   Huber loss.
 
-  ## Argumet Shapes
+  ## Argument Shapes
 
     * `y_true` - $(d_0, d_1, ..., d_n)$
     * `y_pred` - $(d_0, d_1, ..., d_n)$

mix.exs

@@ -116,7 +116,7 @@ defmodule Axon.MixProject do
       groups_for_extras: [
         "Guides: Model Creation": Path.wildcard("guides/model_creation/*.livemd"),
         "Guides: Model Execution": Path.wildcard("guides/model_execution/*.livemd"),
-        "Guides: Training and Evalutaion":
+        "Guides: Training and Evaluation":
           Path.wildcard("guides/training_and_evaluation/*.livemd"),
         "Guides: Serialization": Path.wildcard("guides/serialization/*.livemd"),
         "Examples: Basics": Path.wildcard("notebooks/basics/*.livemd"),
@@ -156,7 +156,7 @@ defmodule Axon.MixProject do
           Axon.MixedPrecision,
           Axon.None,
           Axon.StatefulOutput,
-          Axon.Initalizers
+          Axon.Initializers
         ],
         Summary: [
           Axon.Display

test/axon/compiler_test.exs

@@ -4901,7 +4901,7 @@ defmodule CompilerTest do
       assert Nx.type(b) == {:f, 32}
     end
 
-    test "initializes correclty with single namespace no params" do
+    test "initializes correctly with single namespace no params" do
       model = Axon.input("input_0", shape: {nil, 1}) |> Axon.namespace("model")
 
       {init_fn, _} = Axon.build(model)

test/axon/loop_test.exs

@@ -203,7 +203,7 @@ defmodule Axon.LoopTest do
              end) =~ "Batch"
     end
 
-    test "eval_step/1 evalutes model on a single batch" do
+    test "eval_step/1 evaluates model on a single batch" do
       inp = Nx.tensor([0, 1, 0, 1, 0, 1]) |> Nx.new_axis(-1)
       tar = Nx.tensor([1, 0, 1, 0, 1, 0]) |> Nx.new_axis(-1)