README: document tokenized shards (1 val + 133 train = 13.3B tokens) alongside .model

add SP16384 tokenized shards (13.3B tokens uint16: 1 val + 133 train)

add README with training config + usage

add fineweb_16384_bpe.vocab (SP16384 BPE, docs_selected rev 9bb295dd, sp_seed=1337, 5M train docs, SP 0.2.1, shuffle_input_sentence=False)

add fineweb_16384_bpe.model (SP16384 BPE, docs_selected rev 9bb295dd, sp_seed=1337, 5M train docs, SP 0.2.1, shuffle_input_sentence=False)

initial commit