Suppress benign camembert/roberta model type warning in ThEnTranslator

CITATION.cff

@@ -23,6 +23,7 @@ authors:
 abstract: PyThaiNLP is a Thai natural language processing library for Python. It provides standard linguistic analysis for the Thai language, including tokenization and part-of-speech tagging. Additionally, it offers standard Thai locale utility functions, such as Thai Buddhist Era date formatting and the conversion of numbers into Thai text.
 repository-code: "https://github.com/PyThaiNLP/pythainlp"
 type: software
+doi: 10.5281/zenodo.3519354
 version: 5.2.0
 license-url: "https://spdx.org/licenses/Apache-2.0"
 keywords:

pythainlp/translate/en_th.py

@@ -11,6 +11,7 @@
 from __future__ import annotations
 
 import os
+import warnings
 
 try:
     from fairseq.models.transformer import TransformerModel
@@ -126,26 +127,33 @@ def __init__(self, use_gpu: bool = False):
         self._model_name = _TH_EN_MODEL_NAME
 
         _download_install(self._model_name)
-        self._model = TransformerModel.from_pretrained(
-            model_name_or_path=_get_translate_path(
-                self._model_name,
-                _TH_EN_FILE_NAME,
-                "models",
-            ),
-            checkpoint_file="checkpoint.pt",
-            data_name_or_path=_get_translate_path(
-                self._model_name,
-                _TH_EN_FILE_NAME,
-                "vocab",
-            ),
-            bpe="sentencepiece",
-            sentencepiece_model=_get_translate_path(
-                self._model_name,
-                _TH_EN_FILE_NAME,
-                "bpe",
-                "spm.th.model",
-            ),
-        )
+        # Suppress model type mismatch warning from transformers
+        # The pre-trained model has camembert config but works fine
+        with warnings.catch_warnings():
+            warnings.filterwarnings(
+                "ignore",
+                message="(?i).*using a model of type .* to instantiate a model of type.*",
+            )
+            self._model = TransformerModel.from_pretrained(
+                model_name_or_path=_get_translate_path(
+                    self._model_name,
+                    _TH_EN_FILE_NAME,
+                    "models",
+                ),
+                checkpoint_file="checkpoint.pt",
+                data_name_or_path=_get_translate_path(
+                    self._model_name,
+                    _TH_EN_FILE_NAME,
+                    "vocab",
+                ),
+                bpe="sentencepiece",
+                sentencepiece_model=_get_translate_path(
+                    self._model_name,
+                    _TH_EN_FILE_NAME,
+                    "bpe",
+                    "spm.th.model",
+                ),
+            )
         if use_gpu:
             self._model.cuda()