train issue RuntimeError: Physical devices cannot be modified after being initialized

[MC1016]

when i run python train.py RuntimeError: Physical devices cannot be modified after being initialized

[ryuzou]

I had the same issue and it could be solved by importing the yolov4 after the set_memory_growth method of tensorflow.

from absl import app, flags, logging
from absl.flags import FLAGS
import os
import shutil
import tensorflow as tf
#from core.yolov4 import YOLO, decode, compute_loss, decode_train    #exclude this code
from core.dataset import Dataset
from core.config import cfg
import numpy as np
from core import utils
from core.utils import freeze_all, unfreeze_all

and this at line 19 to 21

    if len(physical_devices) > 0:
        tf.config.experimental.set_memory_growth(physical_devices[0], True)
    from core.yolov4 import YOLO, decode, compute_loss, decode_train  #inset this code here

Or alternatively, comment out the set_memory_growth method at line 20.

[Manoj-M-97]

I had the same issue and it could be solved by importing the yolov4 after the set_memory_growth method of tensorflow.

from absl import app, flags, logging
from absl.flags import FLAGS
import os
import shutil
import tensorflow as tf
#from core.yolov4 import YOLO, decode, compute_loss, decode_train    #exclude this code
from core.dataset import Dataset
from core.config import cfg
import numpy as np
from core import utils
from core.utils import freeze_all, unfreeze_all

and this at line 19 to 21

    if len(physical_devices) > 0:
        tf.config.experimental.set_memory_growth(physical_devices[0], True)
    from core.yolov4 import YOLO, decode, compute_loss, decode_train  #inset this code here

Or alternatively, comment out the set_memory_growth method at line 20.

The first way worked for me. Thanks 👍

[LeamonLee]

I had the same issue and it could be solved by importing the yolov4 after the set_memory_growth method of tensorflow.

from absl import app, flags, logging
from absl.flags import FLAGS
import os
import shutil
import tensorflow as tf
#from core.yolov4 import YOLO, decode, compute_loss, decode_train    #exclude this code
from core.dataset import Dataset
from core.config import cfg
import numpy as np
from core import utils
from core.utils import freeze_all, unfreeze_all

and this at line 19 to 21

    if len(physical_devices) > 0:
        tf.config.experimental.set_memory_growth(physical_devices[0], True)
    from core.yolov4 import YOLO, decode, compute_loss, decode_train  #inset this code here

Or alternatively, comment out the set_memory_growth method at line 20.

Hi, I didn't have this error in the first place, but all of a sudden this error occurred. Why was that?

[Nafees-060]

I got the same issue, does anyone found the solution?

[Tracy-git]

@Nafees-060 i use his method ,solve this problems ,but dont know why .who can tell me

[maryamxasghari]

@Nafees-060 Got the same issue and @Manoj-M-97 method solve my problem. Thank you