add token filter to registry, fix atom parse for LR

pylon · Apr 2, 2018 · ebddb88 · ebddb88
1 parent ebf79ed
commit ebddb88
Show file tree

Hide file tree

Showing 3 changed files with 3 additions and 2 deletions.
diff --git a/lib/penelope/ml/linear/classifier.ex b/lib/penelope/ml/linear/classifier.ex
@@ -93,7 +93,7 @@ defmodule Penelope.ML.Linear.Classifier do
     model =
       params
       |> Map.new(fn {k, v} -> {String.to_existing_atom(k), v} end)
-      |> Map.update!(:solver, &String.to_existing_atom/1)
+      |> Map.update!(:solver, &String.to_atom/1)
       |> Map.put(:classes, Enum.to_list(0..(length(classes) - 1)))
       |> Map.update!(:coef, fn l -> Enum.map(l, &Vector.from_list/1) end)
       |> Map.update!(:intercept, fn

diff --git a/lib/penelope/ml/registry.ex b/lib/penelope/ml/registry.ex
@@ -15,6 +15,7 @@ defmodule Penelope.ML.Registry do
     feature_merge: Penelope.ML.Feature.MergeFeaturizer,
     context_featurizer: Penelope.ML.Feature.ContextFeaturizer,
     lowercase_preprocessor: Penelope.ML.Text.LowercasePreprocessor,
+    token_filter: Penelope.ML.Text.TokenFilter,
     ptb_tokenizer: Penelope.ML.Text.PTBTokenizer,
     ptb_digit_tokenizer: Penelope.ML.Text.PTBDigitTokenizer,
     count_vectorizer: Penelope.ML.Text.CountVectorizer,

diff --git a/mix.exs b/mix.exs
@@ -5,7 +5,7 @@ defmodule Penelope.Mixfile do
     [
       app: :penelope,
       name: "Penelope",
-      version: "0.4.2",
+      version: "0.4.3",
       elixir: "~> 1.6",
       compilers: ["nif" | Mix.compilers()],
       aliases: [clean: ["clean", "clean.nif"]],