CNN fails "Failed to launch ptxas"

,

I have run a simple DNN without problems. Initially the DNN failed with:
"Attempting to perform BLAS operation using StreamExecutor without BLAS support".

Now I’m applying a CNN to the same (fashion MNIST ) classification problem but I get “Failed to launch ptxas” and ultimately:

Node: 'sequential/dense/MatMul'
Attempting to perform BLAS operation using StreamExecutor without BLAS support
	 [[{{node sequential/dense/MatMul}}]] [Op:__inference_train_function_690]

I have also run an NLP model using an embedding layer and global average pooling, and 16 embedding dimensions so I’m not sure why the CNN fails?

Appreciate any guidance!

@brendonwp,

It seems the problem is due to the allocation of GPU memory. You may set the Tensorflow to dynamically allocate GPU memory using

import tensorflow as tf
gpus = tf.config.list_physical_devices(device_type = 'GPU')
tf.config.experimental.set_memory_growth(gpus[0], True)

Thank you!

1 Like