Update Chinese-Japanese bilingual model path to be espnet/mixdata_svs_visinger2_spkemb_lang_pretrained

server.py

@@ -39,7 +39,7 @@ SYSTEM_PROMPT = """
 
 
 config = argparse.Namespace(
-    model_path="espnet/mixdata_svs_visinger2_spkembed_lang_pretrained",
+    model_path="espnet/mixdata_svs_visinger2_spkemb_lang_pretrained",
     cache_dir="cache",
     device="cuda", # "cpu"
     melody_source="random_generate", # "random_select.take_lyric_continuation"

svs_utils.py

@@ -210,7 +210,7 @@ def svs_inference(answer_text, svs_model, config, **kwargs):
     if config.model_path == "espnet/aceopencpop_svs_visinger2_40singer_pretrain":
         sid = np.array([int(config.speaker)])
         output_dict = svs_model(batch, sids=sid)
-    elif config.model_path == "espnet/mixdata_svs_visinger2_spkembed_lang_pretrained":
+    elif config.model_path == "espnet/mixdata_svs_visinger2_spkemb_lang_pretrained":
         langs = {
             "zh": 2,
             "jp": 1,
@@ -359,7 +359,7 @@ if __name__ == "__main__":
 
     # -------- demo code for generate audio from randomly selected song ---------#
     config = argparse.Namespace(
-        model_path="espnet/mixdata_svs_visinger2_spkembed_lang_pretrained",
+        model_path="espnet/mixdata_svs_visinger2_spkemb_lang_pretrained",
         cache_dir="cache",
         device="cuda", # "cpu"
         melody_source="random_select.touhou", #"random_generate" "random_select.take_lyric_continuation",  "random_select.touhou"

util.py

@@ -61,7 +61,7 @@ def get_tokenizer(model, lang):
             return lambda text: split_pinyin_py(text)
         else:
             raise ValueError(f"Only support Chinese language for {model}")
-    elif model == "espnet/mixdata_svs_visinger2_spkembed_lang_pretrained":
+    elif model == "espnet/mixdata_svs_visinger2_spkemb_lang_pretrained":
         if lang == "zh":
             with open(os.path.join("resource/all_plans.json"), "r") as f:
                 all_plan_dict = json.load(f)
@@ -74,7 +74,7 @@ def get_tokenizer(model, lang):
         else:
             raise ValueError(f"Only support Chinese and Japanese language for {model}")
     else:
-        raise ValueError(f"Only support espnet/aceopencpop_svs_visinger2_40singer_pretrain and espnet/mixdata_svs_visinger2_spkembed_lang_pretrained for now")
+        raise ValueError(f"Only support espnet/aceopencpop_svs_visinger2_40singer_pretrain and espnet/mixdata_svs_visinger2_spkemb_lang_pretrained for now")
 
 
 def get_pinyin(texts):