Issues while running

[chrissunny94]

DNN library is not found.
	 [[{{node model/Conv1/Conv2D}}]] [Op:__inference_predict_function_10189]


Num GPUs Available:  1
loading file ...mobilenetv2_weights.h5...!
Traceback (most recent call last):
  File "/home/<>/Desktop/POINTCLOUD/YOLOv8-3D/demo.py", line 116, in <module>
    prediction = bbox3d_model.predict(patch, verbose = 0)
  File "/home/<>/anaconda3/envs/test2/lib/python3.10/site-packages/keras/src/utils/traceback_utils.py", line 70, in error_handler
    raise e.with_traceback(filtered_tb) from None
  File "/home/<>/anaconda3/envs/test2/lib/python3.10/site-packages/tensorflow/python/eager/execute.py", line 60, in quick_execute
    tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
tensorflow.python.framework.errors_impl.UnimplementedError: Graph execution error:

Detected at node model/Conv1/Conv2D defined at (most recent call last):

[bharath5673]

this is beacause both tensorflow and pytorch tries to acocate one GPU which can lead to conflicts. so try on cpu with new env and run..

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
pip3 install tensorflow-cpu

[bharath5673]

u can use one GPU for tensorflow model and pytorch model at a time if u are good in GPU Memory Management..

Dynamic Memory Allocation:
Both TensorFlow and PyTorch can dynamically allocate GPU memory based on the requirements of the models and operations being performed. This allows for more efficient utilization of the available GPU memory.

import torch
import tensorflow as tf

# Set device to GPU (assuming it's available)
device = torch.device('cuda')

# PyTorch Model
pytorch_model = PyTorchModel().to(device)

# TensorFlow Model
tf_config = tf.compat.v1.ConfigProto()
tf_config.gpu_options.allow_growth = True  # Allow dynamic memory growth
tf_session = tf.compat.v1.Session(config=tf_config)

with tf_session.as_default():
    with tf.device('/GPU:0'):
        tensorflow_model = TensorFlowModel()

# Move TensorFlow model to GPU (if it wasn't created on GPU)
with tf_session.as_default():
    with tf.device('/GPU:0'):
        tensorflow_model = tf.device('/GPU:0')(tensorflow_model)
        tensorflow_model.build(input_shape=(None, ...))  # Specify input shape

# Now, you can use both models on the same GPU

[chrissunny94]

Thank you , Bharath