Card: corpus archives consolidated in-repo; data store removed

README.md

@@ -70,13 +70,13 @@ every bucket coverage-gated, all 60 source revisions pinned in `sources.lock.jso
 
 **Where the data lives**
 - In this repo: `DATA.7z` (raw 240 GB archive) · `Corpus_Cleaned.7z` (cleaned archive) · `corpus/` (metadata).
-- Full uncompressed store: [`LUNA-Tokenizer-Corpus-Artifacts`](https://huggingface.co/ASTERIZER/LUNA-Tokenizer-Corpus-Artifacts) — `clean/` (per-family JSONL) + `corpus_parts/` (20 shards).
+- The full uncompressed corpus (`clean/` per-family JSONL + `corpus_parts/` shards) is archived in `Corpus_Cleaned.7z`.
 - Held-out eval sets per language/code: `heldout/`.
 
 ## Sibling repos
 
 - **`tokenizer/`** (inside this repo) — the frozen, ready-to-load tokenizer artifacts (128K / 64K / 8K).
-- **[LUNA-Tokenizer-Corpus-Artifacts](https://huggingface.co/ASTERIZER/LUNA-Tokenizer-Corpus-Artifacts)** — the full uncompressed corpus store (`clean/` per-family + `corpus_parts/` shards), ~432 GB.
+- **Corpus archives** — `Corpus_Cleaned.7z` (cleaned) and `DATA.7z` (raw) live inside this repo; `corpus/` holds the per-family metadata.
 
 ## License