Could not locate cublasLt64_12.dll. Please make sure it is in your library path!

General Discussion
#1

I installed cuda 11.2, cudnn 8.1 and tensorflow 2.8 by following the windows native installation guide. I was running training on a model and it was training fine with my GPU as intended. I then changed the model from a custom model to a sequential one with the following code

model = Sequential()
    model.add(Conv2D(3, (3, 3), activation='relu', padding='same', input_shape=(540, 960, 2*3)))

    model.compile(optimizer='adam', loss='mse', metrics=['accuracy'])
    model.fit(x=numpy.array(x),
              y=numpy.array(y), epochs=50, batch_size=16, validation_split=0.1)
    model.save("model/interpolation_model2")

However after doing this code I’m getting the following error

Epoch 1/50
Could not locate cublasLt64_12.dll. Please make sure it is in your library path!
#3

@A.G_OSMAN,

To run the program successfully, make sure the cublasLt64_12.dll file is in the program’s library path. If it’s not, manually copy it from the bin subfolder of the CUDA Toolkit installation directory.

Thank you!

#5

@chunduriv
So I checked and the dll file is in the bin folder of the environment I was using so not sure what the issue is, is it meant to be in another folder as well

#7

@A.G_OSMAN,

No, then it looks strange. To help us troubleshoot further, could you run any other basic sequential network on the GPU and let us know if the problem persists?

Thank you!