Fix typos and improve language

lib/bumblebee/audio/whisper_featurizer.ex

@@ -15,7 +15,7 @@ defmodule Bumblebee.Audio.WhisperFeaturizer do
     num_seconds: [
       default: 30,
       doc: """
-      the maximum duration of the audio sequence. This implies that the the maximum length of the
+      the maximum duration of the audio sequence. This implies that the maximum length of the
       input sequence is `:num_seconds` * `:sampling_rate`
       """
     ],

lib/bumblebee/layers/transformer.ex

@@ -241,7 +241,7 @@ defmodule Bumblebee.Layers.Transformer do
 
       Alternatively a custom 2-arity function may be given. The function
       should add a normalization node to the given Axon node. The function
-      also receives layer name prefix as the second argument.
+      also receives the layer name prefix as the second argument.
 
     * `:block_type` - controls which configuration of the block to use,
       one of:

lib/bumblebee/shared.ex

@@ -26,7 +26,7 @@ defmodule Bumblebee.Shared do
       use_cross_attention: [
         default: false,
         doc:
-          "whether cross-attention layers should be added to the model." <>
+          "whether cross-attention layers should be added to the model. " <>
             "This is only relevant for decoder models"
       ]
     ]

lib/bumblebee/tokenizer.ex

@@ -92,9 +92,9 @@ defmodule Bumblebee.Tokenizer do
   end
 
   @doc """
-  Converts the given token id the corresponding token.
+  Converts the given token id to the corresponding token.
   """
-  @spec token_to_id(t(), token_id()) :: token()
+  @spec id_to_token(t(), token_id()) :: token()
   def id_to_token(%module{} = tokenizer, id) do
     module.id_to_token(tokenizer, id)
   end

lib/bumblebee/vision/blip_featurizer.ex

@@ -58,17 +58,17 @@ defmodule Bumblebee.Vision.BlipFeaturizer do
     images = List.wrap(images)
 
     for image <- images do
-      images =
+      image =
         image
         |> Image.to_batched_tensor()
         |> Nx.as_type(:f32)
         |> Image.normalize_channels(length(featurizer.image_mean))
 
       if featurizer.resize do
         %{height: height, width: width} = featurizer.size
-        NxImage.resize(images, {height, width}, method: featurizer.resize_method)
+        NxImage.resize(image, {height, width}, method: featurizer.resize_method)
       else
-        images
+        image
       end
     end
     |> Nx.concatenate()

lib/bumblebee/vision/vit_featurizer.ex

@@ -58,17 +58,17 @@ defmodule Bumblebee.Vision.VitFeaturizer do
     images = List.wrap(images)
 
     for image <- images do
-      images =
+      image =
         image
         |> Image.to_batched_tensor()
         |> Nx.as_type(:f32)
         |> Image.normalize_channels(length(featurizer.image_mean))
 
       if featurizer.resize do
         %{height: height, width: width} = featurizer.size
-        NxImage.resize(images, {height, width}, method: featurizer.resize_method)
+        NxImage.resize(image, {height, width}, method: featurizer.resize_method)
       else
-        images
+        image
       end
     end
     |> Nx.concatenate()