Fix mypy errors with compact dependencies: handle numpy return types and remove unused type ignores

Co-authored-by: bact <128572+bact@users.noreply.github.com>

Author: Copilot

pythainlp/summarize/keybert.py

@@ -154,7 +154,7 @@ def embed(self, docs: Union[str, list[str]]) -> np.ndarray:
                 [np.array(emb[0]).mean(axis=0) for emb in embs]
             )
 
-        return emb_mean
+        return emb_mean  # type: ignore[no-any-return]
 
 
 def _generate_ngrams(
@@ -223,10 +223,10 @@ def l2_norm(v: np.ndarray) -> np.ndarray:
         )
         if not np.isclose(np.linalg.norm(result, axis=1), 1).all():
             raise ValueError("Cannot normalize a vector to unit vector.")
-        return result
+        return result  # type: ignore[no-any-return]
 
     def cosine_sim(a: np.ndarray, b: np.ndarray) -> np.ndarray:
-        return (np.matmul(a, b.T).T).sum(axis=1)
+        return (np.matmul(a, b.T).T).sum(axis=1)  # type: ignore[no-any-return]
 
     doc_vector = l2_norm(doc_vector)
     word_vectors = l2_norm(word_vectors)

pythainlp/tag/wangchanberta_onnx.py

@@ -88,7 +88,7 @@ def postprocess(self, logits_data: "np.ndarray") -> "np.ndarray":
         maxes = np.max(logits_t, axis=-1, keepdims=True)
         shifted_exp = np.exp(logits_t - maxes)
         scores = shifted_exp / shifted_exp.sum(axis=-1, keepdims=True)
-        return scores
+        return scores  # type: ignore[no-any-return]
 
     def clean_output(
         self, list_text: list[tuple[str, str]]

pythainlp/tokenize/nlpo3.py

@@ -9,7 +9,7 @@
 from typing import TYPE_CHECKING, Any, Optional
 
 if TYPE_CHECKING:
-    from nlpo3 import (  # type: ignore[import-not-found]
+    from nlpo3 import (
         load_dict as nlpo3_load_dict,  # noqa: F401
     )
     from nlpo3 import segment as nlpo3_segment  # noqa: F401
@@ -89,7 +89,7 @@ def load_dict(file_path: str, dict_name: str) -> bool:
     msg, success = nlpo3_load_dict(file_path=file_path, dict_name=dict_name)
     if not success:
         print(msg, file=stderr)
-    return success  # type: ignore[no-any-return]
+    return success
 
 
 def segment(
@@ -127,7 +127,7 @@ def segment(
     if custom_dict == _NLPO3_DEFAULT_DICT_NAME:
         _ensure_default_dict_loaded()
 
-    return nlpo3_segment(  # type: ignore[no-any-return]
+    return nlpo3_segment(
         text=text,
         dict_name=custom_dict,
         safe=safe_mode,

pythainlp/transliterate/core.py

@@ -179,7 +179,7 @@ def transliterate(
     elif engine == "thaig2p_v2":
         from pythainlp.transliterate.thaig2p_v2 import transliterate  # noqa: I001
     elif engine == "umt5_thaig2p":
-        from pythainlp.translate.umt5_thaig2p import transliterate  # type: ignore[import-not-found,no-redef]  # noqa: I001
+        from pythainlp.translate.umt5_thaig2p import transliterate  # type: ignore[import-untyped,no-redef]  # noqa: I001
     else:  # use default engine: "thaig2p"
         from pythainlp.transliterate.thaig2p import transliterate  # noqa: I001
 

pythainlp/transliterate/thai2rom_onnx.py

@@ -76,7 +76,7 @@ def _prepare_sequence_in(self, text: str) -> "np.ndarray":
             else:
                 idxs.append(self._char_to_ix["<UNK>"])
         idxs.append(self._char_to_ix["<end>"])
-        return np.array(idxs)
+        return np.array(idxs)  # type: ignore[no-any-return]
 
     def romanize(self, text: str) -> str:
         """:param str text: Thai text to be romanized
@@ -131,7 +131,7 @@ def __init__(
 
     def create_mask(self, source_seq: "np.ndarray") -> "np.ndarray":
         mask = source_seq != self.pad_idx
-        return mask
+        return mask  # type: ignore[no-any-return]
 
     def run(
         self, source_seq: "np.ndarray", source_seq_len: List[int]
@@ -196,9 +196,9 @@ def run(
             decoder_input = np.array([topi])
 
             if decoder_input == end_token:
-                return outputs[:di]
+                return outputs[:di]  # type: ignore[no-any-return]
 
-        return outputs
+        return outputs  # type: ignore[no-any-return]
 
 
 _THAI_TO_ROM_ONNX: ThaiTransliterator_ONNX = ThaiTransliterator_ONNX()

pythainlp/transliterate/w2p.py

@@ -151,7 +151,7 @@ def _load_variables(self) -> None:
     def _sigmoid(self, x: "np.ndarray") -> "np.ndarray":
         import numpy as np
 
-        return 1 / (1 + np.exp(-x))
+        return 1 / (1 + np.exp(-x))  # type: ignore[no-any-return]
 
     def _grucell(
         self,
@@ -205,7 +205,7 @@ def _gru(
             h = self._grucell(x[:, t, :], h, w_ih, w_hh, b_ih, b_hh)  # (b, h)
             outputs[:, t, ::] = h
 
-        return outputs
+        return outputs  # type: ignore[no-any-return]
 
     def _encode(self, word: str) -> "np.ndarray":
         import numpy as np
@@ -214,7 +214,7 @@ def _encode(self, word: str) -> "np.ndarray":
         x = [self.g2idx.get(char, self.g2idx["<unk>"]) for char in chars]
         x = np.take(self.enc_emb, np.expand_dims(x, 0), axis=0)
 
-        return x
+        return x  # type: ignore[no-any-return]
 
     def _short_word(self, word: str) -> Optional[str]:
         self.word: str = word

pythainlp/ulmfit/core.py

@@ -239,7 +239,7 @@ def document_vector(
     else:
         raise ValueError("Aggregate by mean or sum")
 
-    return res
+    return res  # type: ignore[no-any-return]
 
 
 def merge_wgts(

pythainlp/word_vector/core.py

@@ -306,7 +306,7 @@ def sentence_vectorizer(self, text: str, use_mean: bool = True) -> ndarray:
         len_words = len(words)
 
         if not len_words:
-            return vec
+            return vec  # type: ignore[no-any-return]
 
         for word in words:
             if word == " " and self.model_name == "thai2fit_wv":
@@ -320,4 +320,4 @@ def sentence_vectorizer(self, text: str, use_mean: bool = True) -> ndarray:
         if use_mean:
             vec /= len_words
 
-        return vec
+        return vec  # type: ignore[no-any-return]