During training data preparation it is used as "--per_host_train_bsz=${TRAIN_BSZ}"
During training it is used as "--train_batch_size=${TRAIN_BSZ}" and when calling data_utils.get_input_fn() it is used as "per_host_bsz=FLAGS.train_batch_size // FLAGS.num_hosts,"
Maybe I missed something so get a bit confused on this, wonder if anyone could explain this a bit?
TRAIN_BSZ=64 is used in text8_large_tpu.sh.
During training data preparation it is used as "--per_host_train_bsz=${TRAIN_BSZ}" During training it is used as "--train_batch_size=${TRAIN_BSZ}" and when calling data_utils.get_input_fn() it is used as "per_host_bsz=FLAGS.train_batch_size // FLAGS.num_hosts,"
Maybe I missed something so get a bit confused on this, wonder if anyone could explain this a bit?