Update app.py

app.py

@@ -1,55 +1,40 @@
-import gradio as gr
-from transformers import AutoTokenizer, AutoModelForCausalLM
-from typing import List, Tuple
 import torch
+import gradio as gr
+from transformers import (
+    AutomaticSpeechRecognitionPipeline,
+    WhisperForConditionalGeneration,
+    WhisperTokenizer,
+    WhisperProcessor,
+)
+from peft import PeftModel, PeftConfig
 
+peft_model_id = "Moustapha91/whisper-small-wolof"
+language = "French"
+task = "transcribe"
 
-model_name = "Hawoly18/Adia_Llama3.1"
-
-# Vérifier si un GPU est disponible
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-
-
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name)
-
-def respond(
-    message: str,
-    history: List[Tuple[str, str]],
-    system_message: str,
-    max_tokens: int,
-    temperature: float,
-    top_p: float,
-) -> str:
-    
-    prompt = system_message
-    for user_msg, assistant_msg in history:
-        prompt += f"\nUser: {user_msg}\nAssistant: {assistant_msg}"
-    prompt += f"\nUser: {message}\nAssistant:"
-
-    
-    inputs = tokenizer(prompt, return_tensors="pt")
-    outputs = model.generate(
-        **inputs,
-        max_length=max_tokens,
-        temperature=temperature,
-        top_p=top_p,
-        do_sample=True,
-    )
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True).split("Assistant:")[-1].strip()
-    return response
-
+peft_config = PeftConfig.from_pretrained(peft_model_id)
+model = WhisperForConditionalGeneration.from_pretrained(
+    peft_config.base_model_name_or_path, 
+    device_map="auto"  # On supprime la quantization en 8 bits
+)
 
-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"),  # Fixed syntax error
-        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
-    ],
-    title="Chatbot Interface"
+model = PeftModel.from_pretrained(model, peft_model_id)
+tokenizer = WhisperTokenizer.from_pretrained(peft_config.base_model_name_or_path, language=language, task=task)
+processor = WhisperProcessor.from_pretrained(peft_config.base_model_name_or_path, language=language, task=task)
+feature_extractor = processor.feature_extractor
+forced_decoder_ids = processor.get_decoder_prompt_ids(language=language, task=task)
+pipe = AutomaticSpeechRecognitionPipeline(model=model, tokenizer=tokenizer, feature_extractor=feature_extractor)
+
+def transcribe(audio):
+    text = pipe(audio, generate_kwargs={"forced_decoder_ids": forced_decoder_ids}, max_new_tokens=255)["text"]
+    return text
+
+iface = gr.Interface(
+    fn=transcribe,
+    inputs=gr.Audio(type="filepath"),  # On supprime 'source' pour éviter l'erreur
+    outputs="text",
+    title="PEFT LoRA + Whisper Small Wolof",
+    description="Realtime demo for Wolof speech recognition using `PEFT-LoRA` fine-tuned Whisper Small model.",
 )
 
-if __name__ == "__main__":
-    demo.launch()
+iface.launch(share=True)