run the second steps happen errors

[lkx971216]

when i run the python main.py -r=model_training -dpc=I_V_T.ini -mtc=I_V_T.ini ，

Traceback (most recent call last):
File "main.py", line 59, in
run_model_training(
File "D:\2023.5.6\three\Battery-SoC-Estimation-main\src\model_training\run.py", line 87, in run_model_training
model.fit(
File "C:\Users\likaixiong\battery-soc-estimation-venv\lib\site-packages\tensorflow\python\keras\engine\training_v1.py", line 795, in fit
return func.fit(
File "C:\Users\likaixiong\battery-soc-estimation-venv\lib\site-packages\tensorflow\python\keras\engine\training_arrays_v1.py", line 644, in fit
return fit_loop(
File "C:\Users\likaixiong\battery-soc-estimation-venv\lib\site-packages\wandb\integration\keras\keras.py", line 155, in new_arrays
for cbk in cbks:
TypeError: 'NoneType' object is not iterable

can you help me solving the problem, thanks。

[owlesia]

Hi! It seems like Wandb is messing up with the TensorFlow. Here are three different quick fixes to the issue I can suggest right now:

1. If you do not plan to run hyperparameter tuning, you can uninstall Wandb:

pip3 uninstall wandb

Then in 'main.py' comment out hyperparameter tuning since it uses Wandb package which we do not have anymore. Comment out the following lines:

from src.hyperparameter_tuning.run import run_hyperparameter_tuning

run_hyperparameter_tuning(
        dpc=args.data_processing_config,
        mtc=args.model_training_config,
        htc=args.hyperparameter_tuning_config,
        wandb_user=args.wandb_user,
)

Now you can run the model training step without any issues.

2. If you plan to run hyperparameter tuning eventually then you would need Wandb later so do not uninstall it. Although we do not need to use Wandb here in model training, adding it solves the issue. Add the following lines to the top of the file 'src/model_training/run.py':

import wandb
from wandb.keras import WandbCallback

Then right before we call 'model.fit' initialize Wandb:

wandb.init(project="soc_model_training")

Finally, add a Wandb callback to 'model.fit':

model.fit(
      X_train,
      y_train,
      epochs=json.loads(mtc_values["epochs"]),
      batch_size=json.loads(mtc_values["batch_size"]),
      validation_data=(
          X_val,
          y_val,
      ),
      shuffle=True,
      callbacks=[WandbCallback()],
  )

To use Wandb, you will need to create an account and export your API key (steps 1 and 2 at https://docs.wandb.ai/quickstart).

3. You could also just remove 'validation_data' argument from 'model.fit' in the file 'src/model_training/run.py'. Although that of course is not an ideal solution if you want to track how accuracy on validation data changes while training the model.