declref-declref_symbols-1B

Procedurally generated decl-ref documents — a scoped declare/reference formal language whose trained-model entropy dynamics match language models on code and natural text — as flat uint16 token-id .bin files. Each document is a random program of { } scopes, DECL name value statements (free, high-entropy content) and REF name value statements whose value tokens repeat an earlier declaration — predictable only by attending back to it. Names and values are 1–3 subtoken symbols; references are recency-weighted; keywords arrive in Markov runs. Token ids: 0=OPEN 1=CLOSE 2=DECL 3=REF, then name-parts and value-parts in disjoint blocks; vocab = 1,028.

Grammar parameters

param	value
seq_length (document)	2048
shards (canonical layout)	256

file	split	tokens
train.bin	train	1,000,341,504
val.bin	val	9,961,472

The bin is a concatenation of fixed-length 2048-token documents (num_tokens is a whole multiple of it): read via tokens.reshape(-1, 2048) to train one document per row. Document alignment matters — a window starting mid-document orphans its references. train (seed 0) and val (a disjoint seed) are independent streams of the same grammar; token count = filesize / 2; the metas carry the full config, including the shard layout that fixes the bin's exact content.

Load a bin with the standard Hugging Face downloader:

from huggingface_hub import hf_hub_download
import numpy as np

path = hf_hub_download(repo_id="alexkstern/declref-declref_symbols-1B", filename="train.bin", repo_type="dataset")
tokens = np.memmap(path, dtype="uint16", mode="r")