app/config.py

import os
from dotenv import load_dotenv
import torch
load_dotenv()
HF_ENV_TOKEN_HAGGING_FACE = os.getenv("HF_TOKEN")
HF_ENV_MODEL_NAME = os.getenv("HF_MODEL_NAME")

# 📝 Parámetros para chat
CHAT_PARAM_MAX_LENGTH = 256
CHAT_MAX_NEW_TOKENS = 512
CHAT_PARAM_DO_SAMPLE = True
CHAT_PARAM_TEMPERATURE = 0.5
CHAT_PARAM_TOP_P = 0.8
CHAT_PARAM_TOP_K = 50
CHAT_LENGTH_PENALTY = 0.8
CHAT_REPETITION_PENALTY = 1.2

CHAT_DEVICE_MAP = "cuda" if torch.cuda.is_available() else "cpu"
CHAT_TORCH_DTYPE = "auto"

KWARGS_DO_SAMPLE = False
KWARGS_TEMPERATURE = 0.7
KWARGS_TOP_P = 0.9
KWARGS_TOP_K = 50
KWARGS_MAX_LENGTH = 512
KWARGS_REPETITION_PENALTY = 1.2