CuDNN Launch Failure (tensorflow-gpu/CUDA)

Traceback (most recent call last): File '/home/alex/anaconda3/envs/tensorflow/lib/python3.6/site-packages/tensorflow/python/client/session.py', line 1322, in _do_call return

Solution 1:

I met the same error. The reason for mine is that my GPU does not have enough memory for the process.

Solution 2:

I'm probably a few of years late to be of any help Alex but I've come up on this issue when on Windows with a specific GPU. Don't ask me why but adding

import os
os.environ['CUDA_VISIBLE_DEVICES'] = '/gpu:0'

if you have a single GPU works for me