TensorFlow throws cuDNN/cuBLAS/cuFFT and MessageFactory errors on Kaggle

Question

I am working on a Kaggle notebook where I am trying to import TensorFlow/Keras to build a simple neural network for tabular data.

However, even with Accelerator = None (CPU only), TensorFlow fails during import with multiple CUDA-related warnings and then crashes with a protobuf error.

Here is the minimal reproducible code:

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Input, Dense

model = Sequential([
    Input(shape=(7,)),
    Dense(7, activation='relu'),
    Dense(7, activation='relu'),
    Dense(1, activation='linear')
])

Unable to register cuFFT factory

Unable to register cuDNN factory

Unable to register cuBLAS factory

failed call to cuInit: UNKNOWN ERROR (303)

AttributeError: 'MessageFactory' object has no attribute 'GetPrototype'

enter image description here here is the image of error

Sagar · Accepted Answer · 2025-11-17 10:12:14Z

0

The AttributeError is an incompatibility issue between TensorFlow and a too-new protobuf version. You can fix this by downgrading the protobuf to older versions like 3.20.3 or others and restarting the session, and then use an environment variable to force CPU usage and silence the warnings.

# Run this cell and restart the kernal
!pip install protobuf==3.20.3

# Add these lines before importing TensorFlow
import os
os.environ['CUDA_VISIBLE_DEVICES'] = '-1' 

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Input, Dense

# rest of the code...

answered Nov 17 at 10:12

Sagar

3462 silver badges10 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

TensorFlow throws cuDNN/cuBLAS/cuFFT and MessageFactory errors on Kaggle

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related