Delete app.py

app.py

@@ -1,142 +0,0 @@
-import os
-import torch
-import pandas as pd
-from torch.utils.data import Dataset, DataLoader
-from torch.optim import AdamW
-from transformers import MarianMTModel, MarianTokenizer
-from fastapi import FastAPI
-from pydantic import BaseModel
-import logging
-
-# ---------------- CONFIG ----------------
-logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
-logger = logging.getLogger(__name__)
-
-DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
-MODEL_NAME = "Helsinki-NLP/opus-mt-en-fr"
-FINETUNED_MODEL_PATH = "fine_tuned_model"
-CSV_FILE = "Data_1.csv"  # Can also be a Hugging Face raw URL
-
-# ---------------- DATASET ----------------
-class TitleDataset(Dataset):
-    def __init__(self, dataframe, tokenizer, max_length=128):
-        self.data = dataframe.copy()
-        self.tokenizer = tokenizer
-        self.max_length = max_length
-        self.data.columns = [c.strip().lower().replace(" ", "_") for c in self.data.columns]
-
-    def __len__(self):
-        return len(self.data)
-
-    def __getitem__(self, idx):
-        row = self.data.iloc[idx]
-        text = f"Generate title from: Year: {row['year']} City: {row['city_name']} Description: {row['original_title']}"
-        target = row['title']
-
-        source = self.tokenizer(text, max_length=self.max_length, padding='max_length', truncation=True, return_tensors='pt')
-        labels = self.tokenizer(target, max_length=self.max_length, padding='max_length', truncation=True, return_tensors='pt')['input_ids']
-        labels[labels == self.tokenizer.pad_token_id] = -100
-
-        return {
-            "input_ids": source['input_ids'].squeeze(),
-            "attention_mask": source['attention_mask'].squeeze(),
-            "labels": labels.squeeze()
-        }
-
-# ---------------- TRAIN FUNCTION ----------------
-def train_model(csv_file, epochs=2, batch_size=8, lr=5e-5):
-    logger.info("Loading dataset...")
-    df = pd.read_csv(csv_file)
-    tokenizer = MarianTokenizer.from_pretrained(MODEL_NAME)
-    model = MarianMTModel.from_pretrained(MODEL_NAME).to(DEVICE)
-
-    dataset = TitleDataset(df, tokenizer)
-    dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=True)
-    optimizer = AdamW(model.parameters(), lr=lr)
-
-    logger.info("Starting fine-tuning...")
-    model.train()
-    for epoch in range(epochs):
-        logger.info(f"Epoch {epoch+1}/{epochs}")
-        for batch in dataloader:
-            input_ids = batch['input_ids'].to(DEVICE)
-            attention_mask = batch['attention_mask'].to(DEVICE)
-            labels = batch['labels'].to(DEVICE)
-
-            outputs = model(input_ids=input_ids, attention_mask=attention_mask, labels=labels)
-            loss = outputs.loss
-            loss.backward()
-            optimizer.step()
-            optimizer.zero_grad()
-        logger.info(f"Epoch {epoch+1} finished. Loss: {loss.item():.4f}")
-
-    if not os.path.exists(FINETUNED_MODEL_PATH):
-        os.makedirs(FINETUNED_MODEL_PATH)
-    model.save_pretrained(FINETUNED_MODEL_PATH)
-    tokenizer.save_pretrained(FINETUNED_MODEL_PATH)
-    logger.info(f"Fine-tuned model saved to {FINETUNED_MODEL_PATH}")
-
-# ---------------- LOAD OR TRAIN MODEL ----------------
-if os.path.exists(FINETUNED_MODEL_PATH):
-    logger.info("Loading fine-tuned model...")
-    tokenizer = MarianTokenizer.from_pretrained(FINETUNED_MODEL_PATH)
-    model = MarianMTModel.from_pretrained(FINETUNED_MODEL_PATH).to(DEVICE)
-else:
-    if os.path.exists(CSV_FILE):
-        logger.info("Fine-tuned model not found. Training new model...")
-        train_model(CSV_FILE)
-        tokenizer = MarianTokenizer.from_pretrained(FINETUNED_MODEL_PATH)
-        model = MarianMTModel.from_pretrained(FINETUNED_MODEL_PATH).to(DEVICE)
-    else:
-        logger.info("Fine-tuned model not found and CSV dataset missing. Using base pre-trained model...")
-        tokenizer = MarianTokenizer.from_pretrained(MODEL_NAME)
-        model = MarianMTModel.from_pretrained(MODEL_NAME).to(DEVICE)
-
-model.eval()
-
-# ---------------- FASTAPI ----------------
-app = FastAPI(title="French Title Generator API")
-
-class InputData(BaseModel):
-    year: str
-    city: str
-    original_title: str
-    max_length: int = 50
-    temperature: float = 0.8
-
-@app.get("/")
-def home():
-    return {"message": "French Title Generator API is running"}
-
-@app.post("/generate")
-def generate(data: InputData):
-    try:
-        input_text = f"Generate title from: Year: {data.year} City: {data.city} Description: {data.original_title}"
-        inputs = tokenizer(input_text, return_tensors="pt", padding=True, truncation=True, max_length=128).to(DEVICE)
-
-        with torch.no_grad():
-            outputs = model.generate(
-                inputs['input_ids'],
-                attention_mask=inputs['attention_mask'],
-                max_length=data.max_length,
-                do_sample=True,
-                temperature=data.temperature,
-                num_beams=3,
-                early_stopping=True
-            )
-
-        generated_title = tokenizer.decode(outputs[0], skip_special_tokens=True)
-        for phrase in ["Le titre de", "Année:", "Ville:", "Description:", "[YEAR]", "[CITY]", "[DESC]", "[TITLE]"]:
-            generated_title = generated_title.replace(phrase, "")
-        generated_title = " ".join(generated_title.split()).strip()
-
-        return {"generated_title": generated_title}
-
-    except Exception as e:
-        return {"error": str(e)}
-
-# ---------------- RUN ----------------
-if __name__ == "__main__":
-    from waitress import serve
-    logger.info("Starting API on port 5000...")
-    serve(app, host="0.0.0.0", port=5000)