tokenizer white space: revert to previous behavior
This commit is contained in:
@@ -373,7 +373,7 @@ def get_bag_of_words_indices(bag_of_words_ids_or_paths: List[str]) -> List[List[
|
||||
filepath = id_or_path
|
||||
with open(filepath, "r") as f:
|
||||
words = f.read().split("\n")
|
||||
bow_indices.append([TOKENIZER.encode(word) for word in words])
|
||||
bow_indices.append([TOKENIZER.encode(word, add_prefix_space=True) for word in words])
|
||||
return bow_indices
|
||||
|
||||
|
||||
|
||||
Reference in New Issue
Block a user