None: marshal data too short

[benpaobamingliang1]

Model: "DeepAR"

---

248/248 [==============================] - 50s 186ms/step - loss: 0.3978 - rmse: 0.9683 - val_loss: 0.4870 - val_rmse: 1.0712
11-01 11:08:17 E hypernets.e._experiment.py 106 - ExperimentID:[HyperTS_56148232a2a64822f2b6a0028102d580] - None: marshal data too short
Traceback (most recent call last):
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hypernets\experiment_experiment.py", line 91, in run
y_eval=self.y_eval, **kwargs)
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hypernets\experiment\compete.py", line 1551, in train
raise e
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hypernets\experiment\compete.py", line 1543, in train
**kwargs)
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hypernets\experiment\compete.py", line 1226, in fit_transform
**kwargs)
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hyperts\framework\compete.py", line 287, in build_estimator
estimators = [hyper_model.load_estimator(trial.model_file) for trial in trials]
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hyperts\framework\compete.py", line 287, in
estimators = [hyper_model.load_estimator(trial.model_file) for trial in trials]
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hyperts\hyper_ts.py", line 478, in load_estimator
return HyperTSEstimator._load(model_file, self.mode)
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hyperts\hyper_ts.py", line 398, in _load
model = BaseDeepEstimator.load_model(model_file)
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hyperts\framework\dl_base.py", line 864, in load_model
model = load_model(h, custom_objects=custom_objects)
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\keras\utils\traceback_utils.py", line 70, in error_handler
raise e.with_traceback(filtered_tb) from None
File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\keras\utils\generic_utils.py", line 857, in func_load
code = marshal.loads(raw_code)
EOFError: marshal data too short
请问为什么会出现这样的问题呢，我训练时会出现这样的报错

[zhangxjohn]

根据您所提供的信息，无法判断。请提供可以复现以上错误的更多信息。

[benpaobamingliang1]

好的，已QQ发您，谢谢您

[zhangxjohn]

遇到以上问题，建议检查tensorflow和numpy的版本兼容性问题。

[happysaten]

Model: "DeepAR"

248/248 [==============================] - 50s 186ms/step - loss: 0.3978 - rmse: 0.9683 - val_loss: 0.4870 - val_rmse: 1.0712 11-01 11:08:17 E hypernets.e._experiment.py 106 - ExperimentID:[HyperTS_56148232a2a64822f2b6a0028102d580] - None: marshal data too short Traceback (most recent call last): File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hypernets\experiment_experiment.py", line 91, in run y_eval=self.y_eval, **kwargs) File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hypernets\experiment\compete.py", line 1551, in train raise e File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hypernets\experiment\compete.py", line 1543, in train **kwargs) File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hypernets\experiment\compete.py", line 1226, in fit_transform **kwargs) File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hyperts\framework\compete.py", line 287, in build_estimator estimators = [hyper_model.load_estimator(trial.model_file) for trial in trials] File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hyperts\framework\compete.py", line 287, in estimators = [hyper_model.load_estimator(trial.model_file) for trial in trials] File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hyperts\hyper_ts.py", line 478, in load_estimator return HyperTSEstimator._load(model_file, self.mode) File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hyperts\hyper_ts.py", line 398, in _load model = BaseDeepEstimator.load_model(model_file) File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\hyperts\framework\dl_base.py", line 864, in load_model model = load_model(h, custom_objects=custom_objects) File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\keras\utils\traceback_utils.py", line 70, in error_handler raise e.with_traceback(filtered_tb) from None File "D:\Program-SetUp\Anaconda\envs\TF2.3\lib\site-packages\keras\utils\generic_utils.py", line 857, in func_load code = marshal.loads(raw_code) EOFError: marshal data too short 请问为什么会出现这样的问题呢，我训练时会出现这样的报错

同样的问题，请问你解决了吗？我的python版本是3.8，tensorflow版本是2.10.1，numpy是1.2.1
运行以下代码时报错

from hyperts import make_experiment
from hyperts.datasets import load_basic_motions

from sklearn.metrics import f1_score
from sklearn.model_selection import train_test_split

data = load_basic_motions()
train_data, test_data = train_test_split(data, test_size=0.2)

experiment = make_experiment(train_data.copy(),
task='classification',
mode='dl',
tf_gpu_usage_strategy=1,
reward_metric='accuracy',
max_trials=30,
early_stopping_rounds=10)

model = experiment.run()

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

y_pred = model.predict(X_test)
y_proba = model.predict_proba(X_test)

scores = model.evaluate(y_test, y_pred, y_proba=y_proba, metrics=['accuracy', 'auc', f1_score])

print(scores)

[zhangxjohn]

感谢提问！

CSDN有这么解决的：https://blog.csdn.net/weixin_41930058/article/details/124276016

这个错误以前的检查点判断，你的np和tf的版本大概率不匹配，tf现在的更新有些不管np。

我这边使用tf=2.4.0和np=1.19.5没有这种问题。你可以尝试更新一下。