The off-device mirror of the Ember training corpus: what was acquired, what was admitted, and what was tokenized. It exists so that a single disk failure on the training host does not destroy the corpus and every token derived from it. The top level is the stage of the pipeline, and every artifact sits at exactly one stage. tokens/ is keyed by tokenizer digest and then dataset-version digest, so shards produced under a different tokenizer or a different admitted set never share a directory. The shard ledger sits beside the shards it describes, because shards without their ledger are bytes with no provenance. This dataset is not under a single licence. Redistribution obligations vary by…
Independent publisher
Hyun Jun Han
wordingone
Models in Library0
Datasets in Library1
Models on Hugging Face—
Followers—
Compare the published evidence
This profile connects the 1 model and dataset record shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.
Published tasks: Text generation.
Reported licenses: other. Open a record to review its specific license and conditions.
Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.