Update app.py

app.py

@@ -1,64 +1,128 @@
+import os
 import gradio as gr
-from huggingface_hub import InferenceClient
+import numpy as np
+import torch
+import scipy.io.wavfile
+from utils.hparams import create_hparams
+from train import load_model
+from synthesis import generate_speech
+from text import text_to_sequence
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+# Path configurations
+MODEL_DIR = "trained_model"
+MODEL_PATH = os.path.join(MODEL_DIR, "checkpoint_latest.pth.tar")
+CONFIG_PATH = os.path.join(MODEL_DIR, "hparams.yml")
+OUTPUT_PATH = "output.wav"
 
+# Download model if it doesn't exist
+def download_model():
+    if not os.path.exists(MODEL_DIR):
+        os.makedirs(MODEL_DIR)
+    
+    if not os.path.exists(MODEL_PATH):
+        print("Downloading model...")
+        # Add model download code here
+        # For example:
+        # !wget -O MODEL_PATH https://path/to/model
+        raise Exception("You need to download the model checkpoint file and place it in trained_model/checkpoint_latest.pth.tar")
+    
+    if not os.path.exists(CONFIG_PATH):
+        print("Downloading config...")
+        # Add config download code here
+        # For example:
+        # !wget -O CONFIG_PATH https://path/to/config
+        raise Exception("You need to download the hparams.yml file and place it in trained_model/hparams.yml")
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+# Initialize model
+def init_model():
+    try:
+        download_model()
+        
+        hparams = create_hparams(CONFIG_PATH)
+        model = load_model(hparams)
+        model.load_state_dict(torch.load(MODEL_PATH, map_location=torch.device('cpu'))['state_dict'])
+        model.eval()
+        
+        return model, hparams
+    except Exception as e:
+        print(f"Error initializing model: {str(e)}")
+        return None, None
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+# Generate speech
+def synthesize(text, model, hparams):
+    try:
+        sequence = np.array(text_to_sequence(text, ['burmese_cleaners']))[None, :]
+        sequence = torch.autograd.Variable(torch.from_numpy(sequence)).cpu().long()
+        
+        mel_outputs, mel_outputs_postnet, _, alignments = model.inference(sequence)
+        
+        with torch.no_grad():
+            waveform = generate_speech(mel_outputs_postnet, hparams)
+        
+        scipy.io.wavfile.write(OUTPUT_PATH, hparams.sampling_rate, waveform)
+        
+        return OUTPUT_PATH, None
+    except Exception as e:
+        return None, str(e)
 
-    messages.append({"role": "user", "content": message})
+# Gradio interface
+def tts_interface(text):
+    if not text.strip():
+        return None, "Please enter some text."
+    
+    global MODEL, HPARAMS
+    if MODEL is None or HPARAMS is None:
+        MODEL, HPARAMS = init_model()
+    
+    if MODEL is None:
+        return None, "Model could not be initialized. Please check logs."
+    
+    audio_path, error = synthesize(text, MODEL, HPARAMS)
+    
+    if error:
+        return None, f"Error generating speech: {error}"
+    
+    return audio_path, "Speech generated successfully!"
 
-    response = ""
+# Initialize global model variables
+MODEL, HPARAMS = None, None
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
+# Create Gradio interface
+demo = gr.Interface(
+    fn=tts_interface,
+    inputs=[
+        gr.Textbox(
+            lines=3, 
+            placeholder="Enter Burmese text here...", 
+            label="Text"
+        )
+    ],
+    outputs=[
+        gr.Audio(label="Generated Speech"),
+        gr.Textbox(label="Status")
     ],
+    title="Myanmar (Burmese) Text-to-Speech",
+    description="""
+    This is a demo of the Myanmar Text-to-Speech system developed by hpbyte.
+    Enter Burmese text in the box below and click 'Submit' to generate speech.
+    
+    GitHub Repository: https://github.com/hpbyte/myanmar-tts
+    """,
+    allow_flagging="never",
+    examples=[
+        ["မင်္ဂလာပါ"],
+        ["မြန်မာစကားပြောစနစ်ကို ကြိုဆိုပါတယ်"],
+        ["ဒီစနစ်ဟာ မြန်မာစာကို အသံအဖြစ် ပြောင်းပေးနိုင်ပါတယ်"],
+    ]
 )
 
+# Initialize model at startup
+try:
+    MODEL, HPARAMS = init_model()
+    print("Model initialized successfully!")
+except Exception as e:
+    print(f"Error initializing model: {str(e)}")
 
+# Launch the app
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()