Preparing Practice Exam…
Please wait a moment while the system generates the exam questions, answers, and study session.
Please wait a moment while the system generates the exam questions, answers, and study session.
Google Cloud Certified Professional Machine Learning Engineer · Google Cloud
Select one answer.
Written by other learners. Comments can disagree with the marked answer — check the official documentation when in doubt.
it was either B or D but B with TPU but you need to code it to use the parallelism so D is more correct.
The Core Problem: Hardware Bottleneck Your current setup is struggling because of the massive scale of your data (3 million images at 2 GB each = 6 Petabytes of data). A single NVIDIA P100 GPU is a relatively older accelerator with limited memory (16 GB) and throughput. Training on 6 PB of data with one GPU is mathematically bound to take an extremely long time, regardless of how much CPU RAM you have.
tf.distribute.Strategy currently does not support TensorFlow's partitioned variables (where a single variable is split across multiple devices) at this time. Hence leaving the option to move to TPU to accelerate the tranining.