Set English language by default for all the LLM models

tools/who_what_benchmark/tests/test_cli_text.py

@@ -159,7 +159,7 @@ def test_text_verbose():
     assert "## Diff:" in result.stderr
 
 
-def test_text_language_autodetect():
+def test_text_language():
     with tempfile.TemporaryDirectory() as temp_dir:
         temp_file_name = os.path.join(temp_dir, "gt.csv")
         result = run_wwb(
@@ -172,6 +172,8 @@ def test_text_language_autodetect():
                 "2",
                 "--device",
                 "CPU",
+                "--language",
+                "cn",
             ]
         )
         assert result.returncode == 0

tools/who_what_benchmark/whowhatbench/text_evaluator.py

@@ -73,21 +73,6 @@
 }
 
 
-def autodetect_language(model):
-    model2language = {
-        "chatglm": "cn",
-        "qwen2": "cn",
-        "qwen": "cn",
-        "baichuan": "cn",
-        "minicpmv": "cn",
-        "internlm": "cn",
-    }
-
-    if not hasattr(model, "config"):
-        return "en"
-    return model2language.get(model.config.model_type, "en")
-
-
 @register_evaluator(
     "text"
 )
@@ -103,7 +88,7 @@ def __init__(
         max_new_tokens=128,
         crop_question=True,
         num_samples=None,
-        language=None,
+        language="en",
         gen_answer_fn=None,
         generation_config=None,
         generation_config_base=None,
@@ -130,9 +115,6 @@ def __init__(
 
         # Take language from the base model if provided
         self.language = language
-        if self.language is None:
-            if base_model is not None:
-                self.language = autodetect_language(base_model)
 
         if base_model:
             self.gt_data = self._generate_data(
@@ -233,11 +215,6 @@ def default_gen_answer(model, tokenizer, prompt, max_new_tokens, crop_question,
                     data = {"prompts": list(self.test_data)}
                 data = pd.DataFrame.from_dict(data)
         else:
-            if self.language is None:
-                print(
-                    "No language detecting in the base model or ground truth data. Taking language from target model."
-                )
-                self.language = autodetect_language(model)
             data = pd.DataFrame.from_dict(default_data[self.language])
 
         prompt_data = data["prompts"]

tools/who_what_benchmark/whowhatbench/wwb.py

@@ -129,8 +129,8 @@ def parse_args():
         "--language",
         type=str,
         choices=["en", "cn"],
-        default=None,
-        help="Used to select default prompts based on the primary model language, e.g. 'en', 'ch'.",
+        default="en",
+        help="Used to select default prompts based on the primary model language, e.g. 'en', 'cn'.",
     )
     parser.add_argument(
         "--hf",