Buckets:
| """Print generated notes next to their decision (state, question, options, gold) for reading by hand.""" | |
| import json, sys, random | |
| from collections import Counter | |
| sys.path.insert(0, "/workspace/think-decide/src") | |
| from gen_notes import option_texts, serialize | |
| path, n = sys.argv[1], int(sys.argv[2]) if len(sys.argv) > 2 else 12 | |
| notes = [json.loads(l) for l in open(path)] | |
| rows = {} | |
| for l in open("/workspace/data/training-legacy/s1/canonical/train.jsonl"): | |
| r = json.loads(l); rows[(r["case_id"], r["view_id"])] = r | |
| print("reject:", Counter(x["reject"] for x in notes), " tokens: median", sorted(x["n_tokens"] for x in notes)[len(notes)//2], | |
| "max", max(x["n_tokens"] for x in notes)) | |
| for x in random.Random(1).sample(notes, min(n, len(notes))): | |
| r = rows[(x["case_id"], x["view_id"])]; (_, q), = r["questions"].items() | |
| texts, names = option_texts(q) | |
| print("=" * 100); print("[%s | %s | gold=%s | reject=%s]" % (x["source"], x["primitive"], x["gold_name"], x["reject"])) | |
| print("STATE:", serialize(r["state"])[:400].replace("\n", " ")); print("Q:", q["instructions"][:200]) | |
| print("OPTIONS:", "; ".join(names)[:300]); print("NOTE:\n" + x["note"]) | |
Xet Storage Details
- Size:
- 1.17 kB
- Xet hash:
- 3fa5e87c13a1ea2f36b597a531493b123af0c158edba0bdaa7af1775ba0a74b7
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.