Upload app.py

app.py

@@ -1,76 +1,104 @@
-#app.py
-import os            
+# app.py
+# =========================================================================
+# 0. Bootstrap — pip upgrades antes de qualquer import sensível
+# =========================================================================
+import os
 import sys
 import subprocess
-# Forçar versões compatíveis
-subprocess.check_call([sys.executable, "-m", "pip", "install", "--upgrade", "urllib3<2.0", "charset_normalizer<3.4"])
-import warnings  # IMPORT NECESSÁRIO antes de usar
-from requests.packages.urllib3.exceptions import DependencyWarning
-# Ignorar o warning do requests
-warnings.simplefilter('ignore', DependencyWarning)
-# No topo do seu ficheiro Python
-BASE_PATH = os.path.dirname(os.path.abspath(__file__))
-# Forçar o output_dir para um local fixo e absoluto
+
+subprocess.check_call([
+    sys.executable, "-m", "pip", "install", "--quiet", "--upgrade",
+    "urllib3<2.0", "charset_normalizer<3.4"
+])
+
+BASE_PATH      = os.path.dirname(os.path.abspath(__file__))
 FINAL_OUTPUT_DIR = os.path.join(BASE_PATH, "trained_model_output")
-import torch
-import logging
+
+# Suprimir warnings antes de qualquer outro import
+import warnings
+warnings.filterwarnings("ignore", category=UserWarning)
+warnings.filterwarnings("ignore", category=DeprecationWarning)
+try:
+    from requests.packages.urllib3.exceptions import DependencyWarning
+    warnings.simplefilter("ignore", DependencyWarning)
+except ImportError:
+    pass
+
+# Silenciar oneDNN / TF (mesmo sem TensorFlow instalado)
+os.environ["TF_CPP_MIN_LOG_LEVEL"] = "0"
+
+# =========================================================================
+# 1. Stdlib
+# =========================================================================
+import glob
+import inspect
 import json
-import chardet
+import logging
 import math
-import psutil
-import traceback
-import time
-import threading
-import webbrowser
 import platform
-import cpuinfo
 import statistics
-import glob
+import threading
+import time
+import traceback
+
+from contextlib import contextmanager, suppress
+from datetime import datetime, timedelta
+from threading import Thread
+from typing import Any, Dict, Union
+from warnings import warn as log_warning
+
+# =========================================================================
+# 2. Third-party — sistema / hardware
+# =========================================================================
+import chardet
+import cpuinfo
 import ctypes as ct
-import importlib
-import inspect
+import psutil
+
+# =========================================================================
+# 3. PyTorch
+# =========================================================================
+import torch
 import torch.nn as nn
 
-#import intel_extension_for_pytorch as ipex
+# =========================================================================
+# 4. Flask
+# =========================================================================
+from flask import Flask, jsonify, render_template, request, send_file, send_from_directory
 
-from typing import Dict, Union, Any
-#webview.create_window("train-12-ok.py", "index-6.html") #iterface autonomo
-#webview.start()
-# Desativa os logs de informação do oneDNN (nível 1) e avisos (nível 2)
-os.environ['TF_CPP_MIN_LOG_LEVEL'] = '0'
+# =========================================================================
+# 5. HuggingFace / PEFT
+# =========================================================================
+from datasets import Dataset, load_from_disk
+from peft import LoraConfig, PeftModel, TaskType, get_peft_model
+from transformers import (
+    AutoModelForCausalLM,
+    AutoTokenizer,
+    DataCollatorForLanguageModeling,
+    Trainer,
+    TrainerCallback,
+    TrainerControl,
+    TrainerState,
+    TrainingArguments,
+)
 
+# =========================================================================
+# 6. Local
+# =========================================================================
+from config_manager import (
+    _load_constants_from_file,
+    map_backend_to_frontend,
+    update_python_constants,
+)
 
-#import tensorflow as tf
-#'0' = Mostra todas as mensagens (padrão).
-#'1' = Filtra as mensagens de INFO.
-#'2' = Filtra as mensagens de INFO e WARNING.
-#'3' = Filtra todas as mensagens, incluindo ERROR.
-# Suprimir avisos Python
-warnings.filterwarnings("ignore", category=UserWarning, module='tensorflow')
-warnings.filterwarnings("ignore", category=DeprecationWarning)
-# Ajusta logger do TensorFlow
-#tf.get_logger().setLevel('ERROR')
+# =========================================================================
+# 7. Logging
+# =========================================================================
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s - %(levelname)s - %(message)s"
+)
 
-from datetime import timedelta
-from datetime import datetime
-from contextlib import suppress
-from threading import Thread
-from flask import Flask, render_template, request, jsonify
-from contextlib import contextmanager
-
-# Importações do Hugging Face
-from transformers import AutoTokenizer, AutoModelForCausalLM, TrainingArguments, Trainer
-from transformers import TrainerCallback, TrainerState, TrainerControl
-from datasets import Dataset, load_from_disk # type: ignore
-from peft import LoraConfig, get_peft_model, PeftModel, TaskType # type: ignore
-from transformers import DataCollatorForLanguageModeling
-from flask import send_from_directory
-#------------------------------------------------------
-from warnings import warn as log_warning 
-from config_manager import _load_constants_from_file, map_backend_to_frontend, update_python_constants
-#pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cpu 
-
-logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
 #---------------------
 app = Flask(__name__)
 all_data = []
@@ -1634,6 +1662,18 @@ def dataset_dir():
         listing.append("(pasta ainda não criada)")
     return jsonify({"listing": listing})
 
+#-------------------------------------------
+@app.route("/api/dataset/download/<path:filename>")
+def dataset_download(filename):
+    """Serve ficheiro do directório de output do dataset para download."""
+    import re
+    # Segurança: só permite nomes de ficheiro simples sem path traversal
+    safe_name = os.path.basename(filename)
+    file_path = os.path.join(DATASET_OUT_DIR, safe_name)
+    if not os.path.isfile(file_path):
+        return jsonify({"error": "File not found."}), 404
+    return send_file(file_path, as_attachment=True, download_name=safe_name)
+
 #--------------------------------------------
 @app.route("/api/train_status")
 def train_status():