Describe the bug
running tensorflow/federated tutorial :TFF simulations with accelerators(simulations_with_accelerators.ipynb)
in colab pro gpu backend

1.in section :Before we begin ,run :
pip install --quiet --upgrade tensorflow_federated_nightly

console shows :
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.

2.in section: Single GPU execution,run :
run_federated_training(
client_epochs_per_round=1,
train_batch_size=16,
test_batch_size=128,
cnn_num_blocks=2,
conv_width_multiplier=4,
server_learning_rate=1.0,
client_learning_rate=0.01,
total_rounds=10,
clients_per_round=16,
rounds_per_eval=2,
)

console shows :
UnknownError: 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 cnn-14-4/conv2d/Conv2D}}]]
[[StatefulPartitionedCall_1/StatefulPartitionedCall/while/loop_body_control/_151/_57]]
(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 cnn-14-4/conv2d/Conv2D}}]]

3. I also try other tutorials, in the CPU backend, they are ok, in the GPU backend, The same error as above is displayed.

Environment (please complete the following information):

• colab pro GPU backend

Expected behavior
running without errors