While monitoring your model training’s GPU utilization, you discover that you have a native synchronous implementation. The training data is split into multiple files. You want to reduce the execution time of your input pipeline. What should you do?
While monitoring your model training’s GPU utilization, you discover that you have a native synchronous implementation. The training data is split into multiple files. You want to reduce the execution time of your input pipeline. What should you do?
To reduce the execution time of your input pipeline when dealing with a native synchronous implementation and multiple files, the best approach is to add parallel interleave to the pipeline. This will allow multiple data files to be read and processed concurrently, making efficient use of available resources and reducing I/O wait times, thereby speeding up the overall pipeline.
It's D https://www.tensorflow.org/guide/data_performance
I think it's D
Multiple files reduce execution time through papalism
D for me
Went with D
"training data split into multiple files", "reduce the execution time of your input pipeline" -> Parallel interleave
D) Parallelisation required