Actions you must take in order to run your ML Data fast
Jun 18, 2021
- Use the
prefetchtransformation to overlap the work of a producer and consumer - Parallelize the data reading transformation using the
interleavetransformation - Parallelize the
maptransformation by setting thenum_parallel_callsargument - Use the
cachetransformation to cache data in memory during the first epoch - Vectorize user-defined functions passed in to the
maptransformation - Reduce memory usage when applying the
interleave,prefetch, andshuffletransformations
