Update app.py

app.py

@@ -1,59 +1,72 @@
+import os
+
+os.environ["TORCHDYNAMO_DISABLE"] = "1"
+os.environ["TORCH_COMPILE_DISABLE"] = "1"
+os.environ["PYTORCH_DISABLE_CUDNN_BENCHMARK"] = "1"
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
+
+import torch
 import gradio as gr
 import numpy as np
-import os
 import spaces
-import sys
+import logging
 from huggingface_hub import login
 
 
+torch._dynamo.config.disable = True
+torch._dynamo.config.suppress_errors = True
+
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
 hf_token = os.getenv("HF_TOKEN")
 if hf_token:
     login(token=hf_token)
 
 
-try:
-    from maliba_ai.tts.inference import BambaraTTSInference
-    from maliba_ai.config.speakers import Adame, Moussa, Bourama, Modibo, Seydou
-    
+tts_model = None
+speakers_dict = None
+model_initialized = False
 
-    print("Loading Bambara TTS model...")
-    tts = BambaraTTSInference()
-    print("Model loaded successfully!")
+@spaces.GPU()
+def initialize_model():
+    """Initialize the TTS model and speakers - called once with GPU context"""
+    global tts_model, speakers_dict, model_initialized
     
-    MODEL_LOADED = True
-except Exception as e:
-    print(f"Error loading model: {e}")
-    MODEL_LOADED = False
-    tts = None
-
+    if not model_initialized:
+        logger.info("Initializing Bambara TTS model...")
+        
+        try:
+            from maliba_ai.tts.inference import BambaraTTSInference
+            from maliba_ai.config.speakers import Adame, Moussa, Bourama, Modibo, Seydou
+            
 
-if MODEL_LOADED:
-    SPEAKERS = {
-        "Adame": Adame,
-        "Moussa": Moussa, 
-        "Bourama": Bourama,
-        "Modibo": Modibo,
-        "Seydou": Seydou
-    }
-else:
-    SPEAKERS = {
-        "Adame": "Adame",
-        "Moussa": "Moussa", 
-        "Bourama": "Bourama",
-        "Modibo": "Modibo",
-        "Seydou": "Seydou"
-    }
+            tts_model = BambaraTTSInference()
+            
+            speakers_dict = {
+                "Adame": Adame,
+                "Moussa": Moussa, 
+                "Bourama": Bourama,
+                "Modibo": Modibo,
+                "Seydou": Seydou
+            }
+            
+            model_initialized = True
+            logger.info("Model initialized successfully!")
+            
+        except Exception as e:
+            logger.error(f"Failed to initialize model: {e}")
+            raise e
+    
+    return tts_model, speakers_dict
 
 def validate_inputs(text, temperature, top_k, top_p, max_tokens):
-    """Validate user inputs"""
     if not text or not text.strip():
         return False, "Please enter some Bambara text."
     
-    if len(text.strip()) > 1000:
-        return False, "Text is too long. Please use shorter text (max 1000 characters)."
-    
-    if not (0.1 <= temperature <= 2.0):
-        return False, "Temperature must be between 0.1 and 2.0"
+    if not (0.001 <= temperature <= 2.0):
+        return False, "Temperature must be between 0.001 and 2.0"
     
     if not (1 <= top_k <= 100):
         return False, "Top-K must be between 1 and 100"
@@ -61,24 +74,20 @@ def validate_inputs(text, temperature, top_k, top_p, max_tokens):
     if not (0.1 <= top_p <= 1.0):
         return False, "Top-P must be between 0.1 and 1.0"
     
-    if not (256 <= max_tokens <= 4096):
-        return False, "Max tokens must be between 256 and 4096"
-    
     return True, ""
 
 @spaces.GPU()
 def generate_speech(text, speaker_name, use_advanced, temperature, top_k, top_p, max_tokens):
-    """Generate speech from Bambara text"""
-    
-    if not MODEL_LOADED:
-        return None, "❌ Model not loaded. Please check the logs for errors."
+
     
     if not text.strip():
         return None, "Please enter some Bambara text."
     
     try:
 
-        speaker = SPEAKERS[speaker_name]
+        tts, speakers = initialize_model()
+        
+        speaker = speakers[speaker_name]
         
         if use_advanced:
             is_valid, error_msg = validate_inputs(text, temperature, top_k, top_p, max_tokens)
@@ -94,7 +103,6 @@ def generate_speech(text, speaker_name, use_advanced, temperature, top_k, top_p,
                 max_new_audio_tokens=int(max_tokens)
             )
         else:
-
             waveform = tts.generate_speech(
                 text=text.strip(),
                 speaker_id=speaker
@@ -107,9 +115,12 @@ def generate_speech(text, speaker_name, use_advanced, temperature, top_k, top_p,
         return (sample_rate, waveform), f"✅ Audio generated successfully"
         
     except Exception as e:
+        logger.error(f"Speech generation failed: {e}")
         return None, f"❌ Error: {str(e)}"
 
 
+SPEAKER_NAMES = ["Adame", "Moussa", "Bourama", "Modibo", "Seydou"]
+
 examples = [
     ["Aw ni ce", "Adame"],
     ["Mali bɛna diya kɔsɛbɛ,  ka a da a kan baara bɛ ka kɛ.", "Moussa"],
@@ -123,159 +134,157 @@ examples = [
     
 ]
 
-
-with gr.Blocks(title="Bambara TTS - EXPERIMENTAL", theme=gr.themes.Soft()) as demo:
-    gr.Markdown("""
-    # 🎤 Bambara Text-to-Speech ⚠️ EXPERIMENTAL
-    
-    Convert Bambara text to speech using AI. This model is currently experimental.
+def build_interface():
+    """Build the Gradio interface for Bambara TTS"""
     
-    **Bambara** is spoken by millions of people in Mali and West Africa.
-    """)
-    
-    with gr.Row():
-        with gr.Column(scale=2):
-            # Input section
-            text_input = gr.Textbox(
-                label="📝 Bambara Text",
-                placeholder="Type your Bambara text here...",
-                lines=3,
-                max_lines=6,
-                value="Aw ni ce"
-            )
-            
-            # Speaker selection
-            speaker_dropdown = gr.Dropdown(
-                choices=list(SPEAKERS.keys()),
-                value="Adame",
-                label="🗣️ Speaker Voice"
-            )
-            
-            # Generation button
-            generate_btn = gr.Button("🎵 Generate Speech", variant="primary", size="lg")
-            
-        with gr.Column(scale=1):
-            # Advanced settings toggle
-            use_advanced = gr.Checkbox(
-                label="⚙️ Use Advanced Settings", 
-                value=False,
-                info="Enable to customize generation parameters"
-            )
-            
-            # Advanced settings (hidden by default)
-            with gr.Group(visible=False) as advanced_group:
-                gr.Markdown("**Advanced Parameters:**")
-                
-                temperature = gr.Slider(
-                    minimum=0.1, 
-                    maximum=2.0, 
-                    value=0.8, 
-                    step=0.1,
-                    label="Temperature",
-                    info="Higher = more varied"
+    with gr.Blocks(title="Bambara TTS - EXPERIMENTAL") as demo:
+        gr.Markdown("""
+        # 🎤 Bambara Text-to-Speech ⚠️ EXPERIMENTAL
+        
+        **Powered by MALIBA-AI**
+        
+        Convert Bambara text to speech. This model is currently experimental.
+        
+        **Bambara** is spoken by millions of people in Mali and West Africa.
+        .
+        """)
+
+
+        
+        with gr.Row():
+            with gr.Column(scale=2):
+                text_input = gr.Textbox(
+                    label="📝 Bambara Text",
+                    placeholder="Type your Bambara text here...",
+                    lines=3,
+                    max_lines=10,
+                    value="I ni ce"
                 )
                 
-                top_k = gr.Slider(
-                    minimum=1, 
-                    maximum=100, 
-                    value=50, 
-                    step=5,
-                    label="Top-K"
+                speaker_dropdown = gr.Dropdown(
+                    choices=SPEAKER_NAMES,
+                    value="Adame",
+                    label="🗣️ Speaker Voice"
                 )
                 
-                top_p = gr.Slider(
-                    minimum=0.1, 
-                    maximum=1.0, 
-                    value=0.9, 
-                    step=0.05,
-                    label="Top-P"
-                )
+                generate_btn = gr.Button("🎵 Generate Speech", variant="primary", size="lg")
                 
-                max_tokens = gr.Slider(
-                    minimum=256, 
-                    maximum=4096, 
-                    value=2048, 
-                    step=256,
-                    label="Max Length"
+            with gr.Column(scale=1):
+                use_advanced = gr.Checkbox(
+                    label="⚙️ Use Advanced Settings", 
+                    value=False,
+                    info="Enable to customize generation parameters"
                 )
-    
-    # Output section
-    gr.Markdown("### 🔊 Generated Audio")
-    
-    audio_output = gr.Audio(
-        label="Generated Speech",
-        type="numpy",
-        interactive=False
-    )
+                
+                with gr.Group(visible=False) as advanced_group:
+                    gr.Markdown("**Advanced Parameters:**")
+                    
+                    temperature = gr.Slider(
+                        minimum=0.1, 
+                        maximum=2.0, 
+                        value=0.8, 
+                        step=0.1,
+                        label="Temperature",
+                        info="Higher = more varied"
+                    )
+                    
+                    top_k = gr.Slider(
+                        minimum=1, 
+                        maximum=100, 
+                        value=50, 
+                        step=5,
+                        label="Top-K"
+                    )
+                    
+                    top_p = gr.Slider(
+                        minimum=0.1, 
+                        maximum=1.0, 
+                        value=0.9, 
+                        step=0.05,
+                        label="Top-P"
+                    )
+                    
+                    max_tokens = gr.Slider(
+                        minimum=256, 
+                        maximum=4096, 
+                        value=2048, 
+                        step=256,
+                        label="Max Length"
+                    )
         
-    status_output = gr.Textbox(
-        label="Status",
-        interactive=False,
-        show_label=False,
-        container=False
-    )
-    
-    # Examples section
-    with gr.Accordion("📚 Try These Examples", open=True):
-        def load_example(text, speaker):
-            return text, speaker, False, 0.8, 50, 0.9, 2048
+        gr.Markdown("### 🔊 Generated Audio")
         
-        gr.Markdown("**Click any example below:**")
-        example_buttons = []
+        audio_output = gr.Audio(
+            label="Generated Speech",
+            type="numpy",
+            interactive=False
+        )
+            
+        status_output = gr.Textbox(
+            label="Status",
+            interactive=False,
+            show_label=False,
+            container=False
+        )
         
-        for i, (text, speaker) in enumerate(examples):
-            btn = gr.Button(f"🎯 {text[:30]}{'...' if len(text) > 30 else ''}", size="sm")
-            btn.click(
-                fn=lambda t=text, s=speaker: load_example(t, s),
-                outputs=[text_input, speaker_dropdown, use_advanced, temperature, top_k, top_p, max_tokens]
-            )
-    
-    # Information section
-    with gr.Accordion("ℹ️ About", open=False):
-        gr.Markdown("""
-        **⚠️ This is an experimental Bambara TTS model.**
+        with gr.Accordion("Try These Examples", open=True):
+            def load_example(text, speaker):
+                return text, speaker, False, 0.8, 50, 0.9, 2048
+            
+            gr.Markdown("**Click any example below:**")
+            
+            for i, (text, speaker) in enumerate(examples):
+                btn = gr.Button(f" {text[:30]}{'...' if len(text) > 30 else ''}", size="sm")
+                btn.click(
+                    fn=lambda t=text, s=speaker: load_example(t, s),
+                    outputs=[text_input, speaker_dropdown, use_advanced, temperature, top_k, top_p, max_tokens]
+                )
         
-        **Common Bambara Phrases:**
-        - **Aw ni ce** - Hello (formal)
-        - **I ni ce** - Hello (informal)  
-        - **I ka kene wa?** - How are you?
-        - **Aw ni tile** - Good afternoon
+        with gr.Accordion(" About", open=False):
+            gr.Markdown("""
+            **⚠️ This is an experimental Bambara TTS model.**
+            - **Languages**: Bambara (bm)  
+            - **Speakers**: 5 different voice options
+            - **Sample Rate**: 16kHz
+            """)
         
-        **Available Speakers:** Adame, Moussa, Bourama, Modibo, Seydou
+        def toggle_advanced(use_adv):
+            return gr.Group(visible=use_adv)
         
-        **Tips:**
-        - Start with default settings
-        - Use shorter texts for better results
-        - Try different speakers for variety
-        """)
-    
-    # Toggle advanced settings visibility
-    def toggle_advanced(use_adv):
-        return gr.Group(visible=use_adv)
+        use_advanced.change(
+            fn=toggle_advanced,
+            inputs=[use_advanced],
+            outputs=[advanced_group]
+        )
+        
+        generate_btn.click(
+            fn=generate_speech,
+            inputs=[text_input, speaker_dropdown, use_advanced, temperature, top_k, top_p, max_tokens],
+            outputs=[audio_output, status_output],
+            show_progress=True
+        )
+        
+        text_input.submit(
+            fn=generate_speech,
+            inputs=[text_input, speaker_dropdown, use_advanced, temperature, top_k, top_p, max_tokens],
+            outputs=[audio_output, status_output],
+            show_progress=True
+        )
     
-    use_advanced.change(
-        fn=toggle_advanced,
-        inputs=[use_advanced],
-        outputs=[advanced_group]
-    )
+    return demo
+
+def main():
+    """Main function to launch the Gradio interface"""
+    logger.info("Starting Bambara TTS Gradio interface.")
     
-    # Wire up the interface
-    generate_btn.click(
-        fn=generate_speech,
-        inputs=[text_input, speaker_dropdown, use_advanced, temperature, top_k, top_p, max_tokens],
-        outputs=[audio_output, status_output]
+    interface = build_interface()
+    interface.launch(
+        server_name="0.0.0.0",
+        server_port=7860,
+        share=False
     )
     
-    # Auto-generate on Enter key in text input
-    text_input.submit(
-        fn=generate_speech,
-        inputs=[text_input, speaker_dropdown, use_advanced, temperature, top_k, top_p, max_tokens],
-        outputs=[audio_output, status_output]
-    )
+    logger.info("Gradio interface launched successfully.")
 
 if __name__ == "__main__":
-    demo.launch(
-        server_name="0.0.0.0",
-        server_port=7860,
-        share=False
-    )
+    main()