It takes like 270 seconds per 10 train steps,
while the output example shows 2-5 seconds max.
Is it because we do not have a GPU\TPU support when running on the Coursera
notebook?
I got this warning when create an array using trax.fastmath.numpy:
WARNING:absl:No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)
Also, this cell should take about 20 seconds according to the comment,
but it takes forever. Anyone else met this problem?? Please respond!!!
@Ron_Ad Please look at the function tweet_to_tensor. Calling get_vocab to get the value for unk_ID is slowing things for you.
Ohh…that was something that I easily missed (by me understanding the term "not hardcode ‘UNK’ "),
Thanks!
