Although I seem to have installed the right versions of tensorflow, cudatoolkit, cudnn and have the right hardware, adequate functions are not found in the .dll files

Here is my test code simple_model.py which I can’t run:

    import tensorflow as tf
    from tensorflow.keras.layers import Conv2D
    from tensorflow.keras.models import Sequential
    
    model = Sequential([Conv2D(3, 3, input_shape=(None, None, 3))])
    
    with tf.device('gpu:0'):
        x = tf.ones((32, 20, 20, 3))
        y = model.predict(x)

here are the command lines I’ve executed:

conda create -n testenv python=3.8.6

conda activate testenv

conda install cudatoolkit=11.2 cudnn=8.1.0

pip install tensorflow==2.9

python N:pathtosimple_model.py

I have a NVIDIA Quadro RTX 6000 GPU with Driver Version: 460.89.

I get the following errors:

2022-07-28 14:25:55.370989: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2022-07-28 14:26:00.539273: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1532] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 22073 MB memory:  -> device: 0, name: Quadro RTX 6000, pci bus id: 0000:65:00.0, compute capability: 7.5
2022-07-28 14:26:10.831136: E tensorflow/stream_executor/cuda/cuda_dnn.cc:389] Could not create cudnn handle: CUDNN_STATUS_ALLOC_FAILED
2022-07-28 14:26:10.831278: W tensorflow/core/framework/op_kernel.cc:1745] OP_REQUIRES failed at conv_ops.cc:1120 : UNIMPLEMENTED: DNN library is not found.
Traceback (most recent call last):
  File "N:pathtosimple_model.py", line 9, in <module>
    y = model.predict(x)
  File "C:pathtominiforge3envstestenvlibsite-packageskerasutilstraceback_utils.py", line 67, in error_handler
    raise e.with_traceback(filtered_tb) from None
  File "C:pathtominiforge3envstestenvlibsite-packagestensorflowpythoneagerexecute.py", line 54, 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 'sequential/conv2d/Conv2D' defined at (most recent call last):
    File "N:pathtosimple_model.py", line 9, in <module>
      y = model.predict(x)

...

    File "C:pathtominiforge3envstestenvlibsite-packageskeraslayersconvolutionalbase_conv.py", line 250, in call
      outputs = self.convolution_op(inputs, self.kernel)
    File "C:pathtominiforge3envstestenvlibsite-packageskeraslayersconvolutionalbase_conv.py", line 225, in convolution_op
      return tf.nn.convolution(
Node: 'sequential/conv2d/Conv2D'
DNN library is not found.
         [[{{node sequential/conv2d/Conv2D}}]] [Op:__inference_predict_function_106]

Can someone please help me understand what is happening??

I’ve also tried without the cudatoolkit and cudnn packages of conda and rather installed CUDA Toolkit and cuDNN on Nvidia’s website, compiled the first one with Visual Studio 2017 but it didn’t change anything.