Merge branch 'dev' into supprot_aysnc_vllm

Author: fengyizhu

ChatTTS/core.py

@@ -273,7 +273,7 @@ def _load(
                 vq_config=asdict(self.config.dvae.vq),
                 dim=self.config.dvae.decoder.idim,
                 coef=coef,
-                device=self.device,
+                device=device,
             )
             .to(device)
             .eval()
@@ -290,8 +290,8 @@ def _load(
             self.config.embed.num_text_tokens,
             self.config.embed.num_vq,
         )
-        embed.from_pretrained(embed_path, device=self.device)
-        self.embed = embed.to(self.device)
+        embed.from_pretrained(embed_path, device=device)
+        self.embed = embed.to(device)
         self.logger.log(logging.INFO, "embed loaded.")
 
         gpt = GPT(
@@ -319,6 +319,7 @@ def _load(
                 decoder_config=asdict(self.config.decoder),
                 dim=self.config.decoder.idim,
                 coef=coef,
+                device=device,
             )
             .to(device)
             .eval()

ChatTTS/model/dvae.py

@@ -179,7 +179,7 @@ def __init__(
         hop_length=256,
         n_mels=100,
         padding: Literal["center", "same"] = "center",
-        device: torch.device = torch.device("cuda"),
+        device: torch.device = torch.device("cpu"),
     ):
         super().__init__()
         self.device = device
@@ -213,7 +213,7 @@ def __init__(
         vq_config: Optional[dict] = None,
         dim=512,
         coef: Optional[str] = None,
-        device: torch.device = torch.device("cuda"),
+        device: torch.device = torch.device("cpu"),
     ):
         super().__init__()
         if coef is None:

examples/ipynb/colab.ipynb

@@ -355,7 +355,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from tools.audio import load_audio\n",
+    "from ChatTTS.tools.audio import load_audio\n",
     "\n",
     "spk_smp = chat.sample_audio_speaker(load_audio(\"sample.mp3\", 24000))\n",
     "print(spk_smp)  # save it in order to load the speaker without sample audio next time\n",