Merge pull request #62 from farthur/master

Debug Dummifier transform and add parameters seq_len to Bert model
MAIF · Oct 7, 2020 · baf644f · baf644f
2 parents a876d32 + 14240ca
commit baf644f
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 1 deletion.
diff --git a/melusine/models/train.py b/melusine/models/train.py
@@ -422,7 +422,7 @@ def _prepare_bert_sequences(self, X):
         else:
             sequence = X["clean_text"].values.tolist()
         seqs = self.tokenizer.batch_encode_plus(
-            sequence, max_length=100, pad_to_max_length=True
+            sequence, max_length=self.seq_size, pad_to_max_length=True
         )
 
         return (

diff --git a/melusine/prepare_email/metadata_engineering.py b/melusine/prepare_email/metadata_engineering.py
@@ -246,4 +246,5 @@ def transform(self, X, y=None):
             X_ = X_.T.reindex(self.dummy_features).T.fillna(0)
             return X_[self.dummy_features].to_dict(orient="records")[0]
         else:
+            X_ = X_.T.reindex(self.dummy_features).T.fillna(0)
             return X_[self.dummy_features]