指定mode为dl会报错

[Bang0518]

采用官方快速开始示例，将代码块

experiment = make_experiment(train_data=train_data.copy(),
                            task='forecast',
                            timestamp='TimeStamp',
                            covariables=['HourSin', 'WeekCos', 'CBWD'])
model = experiment.run()

修改为

experiment = make_experiment(train_data=train_data.copy(),
                            task='forecast',
                            timestamp='TimeStamp',
                            mode = 'dl',
                            covariables=['HourSin', 'WeekCos', 'CBWD'])
model = experiment.run()

将会导致报错如下：

03-22 02:18:25 E hypernets.e._experiment.py 107 - ExperimentID:[HyperTS_a7bae5a247591977073d99a7cb31a213] - None: Not found available trial, change experiment settings and try again pls.
Traceback (most recent call last):
  File "/home/codespace/.local/lib/python3.10/site-packages/hypernets/experiment/_experiment.py", line 91, in run
    model = self.train(self.hyper_model, self.X_train, self.y_train, self.X_test, X_eval=self.X_eval,
  File "/home/codespace/.local/lib/python3.10/site-packages/hypernets/experiment/compete.py", line 1582, in train
    raise e
  File "/home/codespace/.local/lib/python3.10/site-packages/hypernets/experiment/compete.py", line 1573, in train
    step.fit_transform(hyper_model, X_train, y_train, X_test=X_test, X_eval=X_eval, y_eval=y_eval,
  File "/home/codespace/.local/lib/python3.10/site-packages/hypernets/experiment/compete.py", line 1004, in fit_transform
    raise RuntimeError('Not found available trial, change experiment settings and try again pls.')
RuntimeError: Not found available trial, change experiment settings and try again pls.

[Bang0518]

全过程在GitHub的codespace中进行，已确保安装了必要的库

[zhangxjohn]

Thanks for your asking! 是某一次trial出现了，还是一直出现RuntimeError: Not found available trial, change experiment settings and try again pls.最后失败了？

[Bang0518]

是所有trial都出现报错。 在最新版本下，该问题可通过描述进行复现，仅需添加一行

mode = 'dl',

[zhangxjohn]

安装tensorflow了吗？ 我运行所有的CI Test 没有问题。

[Bang0518]

安装tensorflow了吗？ 我运行所有的CI Test 没有问题。

有的，运行前使用了pip install -r requirement.txt 进行的包的安装，使用pip list命令结果如下

[Bang0518]

完整代码如下

from hyperts import make_experiment
from hyperts.datasets import load_network_traffic

from sklearn.model_selection import train_test_split

data = load_network_traffic()
train_data, test_data = train_test_split(data, test_size=0.2, shuffle=False)

model = make_experiment(train_data.copy(),
                        task='multivariate-forecast',
                        mode='dl',
                        timestamp='TimeStamp',
                        covariates=['HourSin', 'WeekCos', 'CBWD']).run()

X_test, y_test = model.split_X_y(test_data.copy())

y_pred = model.predict(X_test)

scores = model.evaluate(y_test, y_pred)

model.plot(forecast=y_pred, actual=test_data)

本地和codespace均可复现该问题

[Bang0518]

如果不指定dl模型，而改用默认的统计学模型，则不会有任何问题

[zhangxjohn]

你试试将tensorflow降到<=2.10.0。我猜测tensorflow的版本（2.16.1）太高了，有些代码被重构了。

[Bang0518]

Collaborator

解决了，我认为有必要把requirement.txt的内容进行修改

[zhangxjohn]

感谢你的建议，后期会进行修正。