Buckets:
| """WO87 phase 2 — document every model's output target convention. | |
| For each index id: load through the reference loader (sha256-verified), | |
| decode a per-language probe, and record the raw output. The recorded | |
| samples are the EVIDENCE for target-accurate renames (the audit found | |
| zero entries documenting their target — renaming before this job would | |
| fabricate). | |
| Usage (a100-large or large-disk flavor, isx-training mounted rw): | |
| python /ckpt/run-087-target-docs/document_targets.py | |
| """ | |
| from __future__ import annotations | |
| import json | |
| import os | |
| from pathlib import Path | |
| OUT = Path("/ckpt") / "run-087-target-docs" | |
| # one canonical probe per language prefix in the zoo | |
| PROBES = { | |
| "ara": "السلام عليكم", | |
| "urd": "استقبال", | |
| "fas": "سلام", | |
| "heb": "שלום", | |
| "tha": "สวัสดี", | |
| "khm": "សួស្តី", | |
| } | |
| def probe_for(model_id: str) -> str: | |
| return PROBES.get(model_id.split("-")[0], "abc") | |
| def main() -> None: | |
| import sys | |
| sys.path.insert(0, "/im/runtime/src") | |
| from interscript_ml import Model | |
| import yaml | |
| OUT.mkdir(parents=True, exist_ok=True) | |
| # resolve ids straight from the shipped index yaml | |
| import urllib.request | |
| index_url = ("https://github.com/interscript/interscript-models/" | |
| "releases/download/index-v12/models-index.yaml") | |
| with urllib.request.urlopen(index_url) as r: | |
| idx = yaml.safe_load(r) | |
| ids = sorted((idx.get("models") or {}).keys()) | |
| print(f"[doc] {len(ids)} ids", flush=True) | |
| doc = {} | |
| for mid in ids: | |
| try: | |
| m = Model.load(mid) | |
| probe = probe_for(mid) | |
| out = m.translate(probe) | |
| doc[mid] = {"probe": probe, "output": out, | |
| "chars": sorted(set(out))} | |
| print(f"[doc] {mid}: {probe!r} -> {out!r}", flush=True) | |
| except Exception as e: | |
| doc[mid] = {"error": repr(e)[:200]} | |
| print(f"[doc] {mid}: ERROR {e!r}"[:160], flush=True) | |
| (OUT / "target-documentation.json").write_text( | |
| json.dumps(doc, ensure_ascii=False, indent=2), encoding="utf-8") | |
| print("[done]", flush=True) | |
| if __name__ == "__main__": | |
| main() | |
Xet Storage Details
- Size:
- 2.2 kB
- Xet hash:
- 4ecb6bc2e0f9d7056810f26b9c5dbe9b015bc88d5e8f58ca80a7df0039f063ca
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.