Delete app.py

app.py

@@ -1,153 +0,0 @@
-import gradio as gr
-import torch
-import os
-import numpy as np
-import soundfile as sf
-import tempfile
-from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
-from peft import PeftModel
-from huggingface_hub import snapshot_download
-
-# ── Config ────────────────────────────────────────────────────────
-BASE_MODEL = "openai/whisper-large-v3-turbo"
-ADAPTER_REPO = "mozed54/whisper-finetuned"
-HF_TOKEN = os.environ.get("HF_TOKEN")
-
-# ── Chargement ────────────────────────────────────────────────────
-device = "cuda" if torch.cuda.is_available() else "cpu"
-torch_dtype = torch.float16 if device == "cuda" else torch.float32
-
-print(f"Device : {device.upper()}")
-print("Chargement du modele de base...")
-
-base_model = AutoModelForSpeechSeq2Seq.from_pretrained(
-    BASE_MODEL,
-    torch_dtype=torch_dtype,
-    use_safetensors=True,
-)
-
-print("Chargement adapter LoRA...")
-adapter_path = snapshot_download(
-    repo_id=ADAPTER_REPO,
-    repo_type="dataset",
-    token=HF_TOKEN,
-)
-model = PeftModel.from_pretrained(base_model, adapter_path)
-model = model.merge_and_unload()
-model.to(device)
-
-processor = AutoProcessor.from_pretrained(BASE_MODEL)
-
-pipe = pipeline(
-    "automatic-speech-recognition",
-    model=model,
-    tokenizer=processor.tokenizer,
-    feature_extractor=processor.feature_extractor,
-    torch_dtype=torch_dtype,
-    device=device,
-    generate_kwargs={"language": "arabic", "task": "transcribe"},
-)
-
-print("Modele pret !")
-
-# ── Fonctions ─────────────────────────────────────────────────────
-def process_audio(audio):
-    """Accepte tuple (sr, array) ou filepath string"""
-    if audio is None:
-        return None
-
-    if isinstance(audio, tuple):
-        sample_rate, data = audio
-        if data.ndim > 1:
-            data = data.mean(axis=1)
-        data = data.astype(np.float32)
-        if np.abs(data).max() > 1.0:
-            data = data / 32768.0
-        tmp = tempfile.NamedTemporaryFile(suffix=".wav", delete=False)
-        sf.write(tmp.name, data, sample_rate)
-        return tmp.name
-
-    return audio  # déjà un filepath
-
-
-def transcribe(audio):
-    path = process_audio(audio)
-    if path is None:
-        return "Aucun audio recu."
-    try:
-        result = pipe(path, return_timestamps=False)
-        return result["text"].strip()
-    except Exception as e:
-        return f"Erreur : {str(e)}"
-
-
-# ── CSS ───────────────────────────────────────────────────────────
-css = """
-@import url("https://fonts.googleapis.com/css2?family=DM+Sans:wght@400;500;600&display=swap");
-body, .gradio-container { font-family: "DM Sans", sans-serif !important; background: #0f0f13 !important; }
-.gradio-container { max-width: 860px !important; margin: 0 auto !important; padding: 2rem 1.5rem !important; }
-#app-header { text-align: center; margin-bottom: 2rem; padding: 2rem 0 1.5rem; border-bottom: 1px solid #1e1e2a; }
-#app-header h1 { font-size: 2rem; font-weight: 600; color: #f0f0f5; margin: 0 0 0.4rem; }
-#app-header p { font-size: 0.9rem; color: #6b6b80; margin: 0; }
-.section-label { font-size: 0.72rem; font-weight: 600; color: #4a4a60; letter-spacing: 0.12em; text-transform: uppercase; margin-bottom: 0.6rem; }
-button.primary { background: #5a5aff !important; color: #fff !important; border: none !important; border-radius: 10px !important; font-weight: 500 !important; padding: 0.7rem 1.4rem !important; transition: background 0.2s !important; }
-button.primary:hover { background: #4848e0 !important; }
-button.secondary { background: #1e1e2a !important; color: #a0a0b8 !important; border: 1px solid #2a2a38 !important; border-radius: 10px !important; padding: 0.7rem 1.4rem !important; }
-#transcription-box textarea { background: #0f0f13 !important; border: 1.5px solid #1e1e2a !important; border-radius: 14px !important; color: #e0e0ec !important; font-size: 1rem !important; line-height: 1.7 !important; padding: 1.2rem !important; direction: rtl !important; text-align: right !important; transition: border-color 0.4s, box-shadow 0.4s !important; min-height: 160px !important; }
-#transcription-box.has-text textarea { border-color: #22c97a !important; box-shadow: 0 0 0 3px rgba(34,201,122,0.12) !important; }
-footer { display: none !important; }
-"""
-
-js = """
-function() {
-    const observer = new MutationObserver(() => {
-        const box = document.querySelector("#transcription-box textarea");
-        const wrapper = document.querySelector("#transcription-box");
-        if (!box || !wrapper) return;
-        if (box.value && box.value.trim().length > 0) { wrapper.classList.add("has-text"); }
-        else { wrapper.classList.remove("has-text"); }
-    });
-    observer.observe(document.body, { childList: true, subtree: true, characterData: true });
-    return [];
-}
-"""
-
-# ── Interface ─────────────────────────────────────────────────────
-with gr.Blocks(title="Whisper Darija", css=css, js=js) as demo:
-
-    gr.HTML('<div id="app-header"><h1>🎙 Whisper · Darija</h1><p>Transcription automatique — dialecte algerien</p></div>')
-
-    gr.HTML('<div class="section-label">Source audio</div>')
-    with gr.Row():
-        with gr.Column(scale=1):
-            mic_input = gr.Audio(
-                sources=["microphone"],
-                type="numpy",
-                label="🎤 Enregistrer ma voix",
-            )
-            mic_btn = gr.Button("Transcrire la voix", variant="primary")
-
-        with gr.Column(scale=1):
-            file_input = gr.Audio(
-                sources=["upload"],
-                type="numpy",
-                label="📂 Charger un fichier audio",
-            )
-            file_btn = gr.Button("Transcrire le fichier", variant="secondary")
-
-    gr.HTML("<hr style='border:none;border-top:1px solid #1e1e2a;margin:1.5rem 0;'>")
-    gr.HTML('<div class="section-label">Transcription</div>')
-
-    output = gr.Textbox(
-        label="",
-        lines=6,
-        placeholder="النص سيظهر هنا",
-        interactive=False,
-        elem_id="transcription-box",
-        show_label=False,
-    )
-
-    mic_btn.click(fn=transcribe, inputs=[mic_input], outputs=[output])
-    file_btn.click(fn=transcribe, inputs=[file_input], outputs=[output])
-
-demo.launch()