updating to Soprano-1.1-80M

app.py

@@ -8,6 +8,36 @@ try:
     import torch  # type: ignore
 except Exception:  # pragma: no cover
     torch = None  # type: ignore
+
+# Patch soprano's TransformersModel to fix dtype bug in v0.2.0
+# The library incorrectly uses 'dtype=' instead of 'torch_dtype=' in from_pretrained()
+def _patch_soprano_transformers():
+    try:
+        from transformers import AutoModelForCausalLM, AutoTokenizer
+        import soprano.backends.transformers as soprano_transformers
+        
+        class PatchedTransformersModel(soprano_transformers.BaseModel):
+            def __init__(self, device='cuda', model_path=None, **kwargs):
+                self.device = device
+                model_name_or_path = model_path if model_path else 'ekwek/Soprano-1.1-80M'
+                self.model = AutoModelForCausalLM.from_pretrained(
+                    model_name_or_path,
+                    torch_dtype=torch.bfloat16 if device == 'cuda' else torch.float32,
+                    device_map=device
+                )
+                self.tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
+                self.model.eval()
+            
+            # Inherit all other methods from original
+            infer = soprano_transformers.TransformersModel.infer
+            stream_infer = soprano_transformers.TransformersModel.stream_infer
+        
+        soprano_transformers.TransformersModel = PatchedTransformersModel
+    except Exception as e:
+        print(f"Warning: Could not patch soprano transformers backend: {e}")
+
+_patch_soprano_transformers()
+
 try:
     from soprano import SopranoTTS  # type: ignore
 except Exception:  # pragma: no cover
@@ -76,7 +106,7 @@ def soprano_tts(
 
 # --- Gradio UI ---
 with gr.Blocks() as demo:
-    gr.HTML("<h1 style='text-align: center;'>Soprano-TTS</h1><p style='text-align: center;'>Powered by Soprano-80M | 32kHz High-Fidelity Audio</p>")
+    gr.HTML("<h1 style='text-align: center;'>Soprano-TTS</h1><p style='text-align: center;'>Powered by Soprano-1.1-80M | 32kHz High-Fidelity Audio</p>")
 
     with gr.Row(variant="panel"):
         temperature = gr.Slider(