minor

requirements.txt

@@ -9,7 +9,6 @@ unidic-lite
 sudachipy 
 sudachidict_full
 spacy
-scikit-learn==1.2.0
 speach==0.1a15.post1
 torch>=1.13.1
 transformers>=4.25.1

yomikata/utils.py

@@ -8,7 +8,6 @@ import unicodedata
 
 import numpy as np
 import pynvml
-from sklearn.metrics import precision_recall_fscore_support
 from speach.ttlig import RubyFrag, RubyToken
 
 """
@@ -30,9 +29,7 @@ def load_dict(filepath: str) -> dict:
     return d
 
 
-def save_dict(
-    d: dict, filepath: str, cls: json.JSONEncoder = None, sortkeys: bool = False
-) -> None:
+def save_dict(d: dict, filepath: str, cls: json.JSONEncoder = None, sortkeys: bool = False) -> None:
     """Save a dictionary to a specified location.
 
     Args:
@@ -276,10 +273,7 @@ def remove_furigana(s: str) -> str:
     """
     rubytoken = parse_furigana(s)
     return "".join(
-        [
-            token.text if isinstance(token, RubyFrag) else token
-            for token in rubytoken.groups
-        ]
+        [token.text if isinstance(token, RubyFrag) else token for token in rubytoken.groups]
     )
 
 
@@ -295,10 +289,7 @@ def furigana_to_kana(s: str) -> str:
     """
     rubytoken = parse_furigana(s)
     return "".join(
-        [
-            token.furi if isinstance(token, RubyFrag) else token
-            for token in rubytoken.groups
-        ]
+        [token.furi if isinstance(token, RubyFrag) else token for token in rubytoken.groups]
     )
 
 
@@ -315,40 +306,3 @@ def has_kanji(s: str) -> bool:
         if code >= UNICODE_KANJI_START and code <= UNICODE_KANJI_END:
             return True
     return False
-
-
-"""
-Performance Metrics
-"""
-
-
-def get_label_performance(y_true, y_pred, classes):
-    """Per-class performance metrics.
-
-    MIT License
-    Copyright (c) 2020 Made With ML
-    """
-
-    # Performance
-    performance = {"overall": {}, "class": {}}
-
-    # Overall performance
-    metrics = precision_recall_fscore_support(y_true, y_pred, average="weighted")
-    performance["overall"]["precision"] = metrics[0]
-    performance["overall"]["recall"] = metrics[1]
-    performance["overall"]["f1"] = metrics[2]
-    performance["overall"]["num_samples"] = np.float64(len(y_true))
-
-    # Per-class performance
-    metrics = precision_recall_fscore_support(
-        y_true, y_pred, average=None, labels=classes
-    )
-    for i in range(len(classes)):
-        performance["class"][classes[i]] = {
-            "precision": metrics[0][i],
-            "recall": metrics[1][i],
-            "f1": metrics[2][i],
-            "num_samples": np.float64(metrics[3][i]),
-        }
-
-    return performance