RuntimeError: Error running the Tensorflow session. 2 root error(s) found.

[GeorgeMcIntire]

I'm trying to run a VGG genre classifier in Google Colab but I keep use it because this error keeps popping up and I don't know what to do. Thanks.

RuntimeError: In TensorflowPredictVGGish.compute: TensorflowPredict: Error running the Tensorflow session. 2 root error(s) found. (0) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above. [[{{node model/1CNN/Conv2D}}]] [[model/Sigmoid/_3]] (1) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above. [[{{node model/1CNN/Conv2D}}]] 0 successful operations. 0 derived errors ignored.

[palonso]

Could you provide a link to the Colab and the CUDA/CuDNN versions you are using?

[GeorgeMcIntire]

Here is the link to the colab

And here's the cuda version:

Cuda compilation tools, release 11.1, V11.1.105 Build cuda_11.1.TC455_06.29190527_0

[palonso]

Essentia was built with CUDA 11.2 and it is incompatible with CUDA 11.1. While I couldn't find an easy way to update the CUDA version in colab, by coping you colab I was assigned a machine with CUDA 11.2 and your code ran smoothly.

Just one note: TensorflowPredictVGGish is reserved for VGGish based models, the VGG ones are intended to be used with TensorflowPredictMusiCNN.