Question

我一直在尝试使用自定义损失函数（SMAPE）实现轻型gbm模型，但是当我运行它时，python崩溃并显示以下错误

python.exe中0x00007FF841D04E65（lib_lightgbm.dll）处未处理的异常：0xC0000005：访问冲突读取位置0x000002624155B500

在没有自定义损失函数且我仅使用内置rmse的情况下，lightgbm模型似乎正在运行。

此错误是什么意思？我该怎么做才能纠正它？感谢您的提前帮助

def smape(preds, target):
  '''
  Function to calculate SMAPE
  '''
  n = len(preds)
  masked_arr = ~(target==0)
  preds, target = preds[masked_arr], target[masked_arr]
  num = np.abs(preds-target)
  denom = np.abs(target)
  smape_val = (100*np.sum(num/denom))/n
  return smape_val

def lgbm_smape(preds, train_data):
  '''
  Custom Evaluation Function for LGBM
  '''
  labels = train_data.get_label()
  smape_val = smape(np.expm1(preds), np.expm1(labels))
  return 'MAPE', smape_val, False

def grads(x, y) : 
  return 2 * y * (x - y) / ((x + y) * (x + y) * abs(x - y))


def hesss(x, y) :
  return (-4 * y * (x - y)) / ((x + y) * (x + y) * (x + y) * abs(x - y)) 

def lgbm_obj(preds, train_data):
  '''
  Custom obj Function for LGBM
  '''
  labels = train_data.get_label()
  masked_arr = ~((preds==0)&(labels==0))
  preds, labels = preds[masked_arr], labels[masked_arr]
  grad = grads(labels,preds)
  hess = hesss(labels,preds)
  return grad, hess

  import lightgbm as lgb
  params = {'task':'train', 
            'boosting_type':'dart', 
            'objective':'regression', 
            'metric': {'rmse'}, 'num_leaves': 100, 'learning_rate': 0.003, 
            'feature_fraction': 0.8, 'max_depth': 6, 'verbose': 0, 
            'num_boost_round':35000, 'nthread':-1
            }
  lgbtrain = lgb.Dataset(data=X_train.values, 
                     label=y_train['demandQuantity'].values)

  lgbval = lgb.Dataset(data=X_test.values, 
                     label=y_test['demandQuantity'].values, 
                      reference=lgbtrain)


  model = lgb.train(params, lgbtrain, 
                    num_boost_round=params['num_boost_round'], 
                    valid_sets=[lgbtrain, lgbval], feval=lgbm_smape,
                    fobj=lgbm_obj,verbose_eval=200
                    )

使用自定义损失功能运行Light gbm时Python崩溃

0 个答案: