app.py

import gradio as gr
from transformers import WhisperForConditionalGeneration, WhisperProcessor
import torch
import librosa
import warnings
import numpy as np

# -------------------------------
# 0. SUPPRESS WARNINGS
# -------------------------------
warnings.filterwarnings("ignore", category=ResourceWarning)
warnings.filterwarnings("ignore", category=FutureWarning)

# -------------------------------
# 1. CONFIGURATION
# -------------------------------
MODEL_PATH = "MaryWambo/whisper-base-kikuyu4"
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Loading model on {device}...")

# -------------------------------
# 2. LOAD MODEL & PROCESSOR
# -------------------------------
processor = WhisperProcessor.from_pretrained(MODEL_PATH)
model = WhisperForConditionalGeneration.from_pretrained(MODEL_PATH).to(device)

# Force transcription mode
model.config.forced_decoder_ids = processor.get_decoder_prompt_ids(
    language="swahili",
    task="transcribe"
)

# -------------------------------
# 3. CUSTOM CSS
# -------------------------------
theme_styles = """
body, .gradio-container { background-color: white !important; }

#title-text h1 {
    color: #8b0000 !important;
    font-weight: 900 !important;
    text-align: center;
}

.upload-button svg, .mic-button svg {
    transform: scale(1.5) !important;
    color: #8b0000 !important;
}

#predict-box textarea {
    font-size: 1.6rem !important;
    font-weight: 800 !important;
    color: #000000 !important;
    border: 3px solid #8b0000 !important;
}

#run-btn {
    background: #8b0000 !important;
    color: white !important;
    font-weight: bold !important;
    font-size: 1.4rem !important;
}
"""

# -------------------------------
# 4. TRANSCRIPTION FUNCTION
# -------------------------------
def transcribe_kikuyu(audio):
    if audio is None:
        return "Please record or upload audio."

    try:
        # Load audio
        speech_array, sr = librosa.load(audio, sr=16000)
        
        # Convert to float32
        if speech_array.dtype != np.float32:
            speech_array = speech_array.astype(np.float32)
        
        # Tokenize
        inputs = processor(speech_array, sampling_rate=sr, return_tensors="pt")
        input_features = inputs.input_features.to(device)

        # Generate transcription
        with torch.no_grad():
            predicted_ids = model.generate(
                input_features,
                num_beams=5,
                max_new_tokens=255
            )

        transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[0]
        return transcription

    except Exception as e:
        return f"Error during transcription: {str(e)}"

# -------------------------------
# 5. GRADIO UI
# -------------------------------
with gr.Blocks() as demo:
    gr.Markdown("# 🎙️ Kikuyu ASR", elem_id="title-text")

    with gr.Row():
        with gr.Column():
            audio_input = gr.Audio(
                sources=["microphone", "upload"],
                type="filepath",
                label="🎤 Record or Upload Kikuyu Speech"
            )

            submit_btn = gr.Button(
                "🚀 RUN TRANSCRIPTION",
                elem_id="run-btn"
            )

        with gr.Column():
            text_out = gr.Textbox(
                label="🤖 AI Prediction",
                elem_id="predict-box",
                lines=8
            )

    submit_btn.click(
        fn=transcribe_kikuyu,
        inputs=audio_input,
        outputs=text_out
    )

# -------------------------------
# 6. LAUNCH APP
# -------------------------------
import asyncio
import sys

def _suppress_event_loop_closed(loop, context):
    if "Invalid file descriptor" in str(context.get("exception", "")):
        return
    loop.default_exception_handler(context)

try:
    loop = asyncio.get_event_loop()
    loop.set_exception_handler(_suppress_event_loop_closed)
except RuntimeError:
    pass

demo.launch(ssr_mode=False)