Buckets:
| import json, re, sys, glob, os | |
| IDX=json.load(open("/root/27B/docs/fable_review_2026-07-02/verifier_pilot/api_index.json")) | |
| S2F=IDX["short_to_full"]; T=IDX["types"] | |
| def tmembers(short): | |
| m=set() | |
| for fn in S2F.get(short,[]): m|=set(T[fn].get("members",[])) | |
| return m | |
| def exists(typ,mem): | |
| ms=tmembers(typ) | |
| if mem in ms: return True | |
| if (mem.startswith(("get_","set_"))) and mem[4:] in ms: return True | |
| if ("get_"+mem) in ms or ("set_"+mem) in ms: return True | |
| return False | |
| MEMBER=re.compile(r'\b([A-Z][A-Za-z0-9]{2,})\.([A-Za-z_][A-Za-z0-9_]*)') | |
| def score(path): | |
| rows=[json.loads(l) for l in open(path)] | |
| known=e=m=0; miss_examples=[] | |
| for r in rows: | |
| for typ,mem in set(MEMBER.findall(r["generation"])): | |
| if typ in S2F: | |
| known+=1 | |
| if exists(typ,mem): e+=1 | |
| else: | |
| m+=1 | |
| if len(miss_examples)<6: miss_examples.append(f"{typ}.{mem}") | |
| fid=e/(e+m) if (e+m) else None | |
| return len(rows),known,e,m,fid,miss_examples | |
| print("%-10s %5s %8s %6s %6s %8s"%("ckpt","rows","revitRef","exist","miss","API-fid")) | |
| for f in sorted(glob.glob("/root/27B/grade_out/gen_*.jsonl")): | |
| n,k,e,m,fid,ex=score(f) | |
| lab=os.path.basename(f).replace("gen_","").replace(".jsonl","") | |
| print("%-10s %5d %8d %6d %6d %8s"%(lab,n,k,e,m,("%.1f%%"%(fid*100)) if fid is not None else "n/a")) | |
Xet Storage Details
- Size:
- 1.4 kB
- Xet hash:
- 28b87db63f6f4ef8e8a134646e8f5b625aafea35cd16cb65948b9fe0af769c83
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.