File size: 6,277 Bytes
ce33d98 14d44a6 ce33d98 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 | """Select candidate MP tasks for the rows under review and extract their full outputs from S3."""
import sys, json, re, gzip, time, numpy as np, pandas as pd, requests
from pathlib import Path
from concurrent.futures import ProcessPoolExecutor, as_completed
W=Path(sys.argv[1]); OUT=W/"s3"; OUT.mkdir(exist_ok=True)
MAN=Path("/home/kna/.local/share/wyformer/cache/mp_forces_recovery")
S3="https://materialsproject-parsed.s3.amazonaws.com/"
TASK_ID_RE=re.compile(rb'"task_id":\s*"([a-z]+-\d+)"')
def legacy(i):
p,_,b=i.rpartition("-")
if not re.fullmatch(r"[a-z]+",b): return i
n=0
for c in b: n=n*26+ord(c)-97
return f"{p or 'mp'}-{n}"
def select():
f=pd.read_parquet(W/"identify.parquet")
core={d["material_id"]:d for d in json.load(open(W/"core_lemat.json"))}
alias=json.load(open(W/"core_alias.json")) if (W/"core_alias.json").exists() else {}
for i,parent in alias.items(): core.setdefault(i,core[parent])
tasks={t["task_id"]:t for t in json.load(open(W/"tasks_meta.json"))}
stale=f.grp.isin(["deprecated","dep_runmismatch","unidentified","runmismatch"])
livebig=(f.grp=="live_other_task")&(f.dE_mp.abs()>0.05)
ctrl1=f[f.grp=="same_task"].sample(300,random_state=0).index
ctrl2=f[(f.grp=="live_other_task")&(f.dE_mp.abs()<=0.05)].sample(300,random_state=0).index
f["review"]=np.where(stale,"stale",np.where(livebig,"live_big",""))
f.loc[ctrl1,"review"]="control_same"; f.loc[ctrl2,"review"]="control_live"
r=f[f.review!=""]
pairs=[]
for row in r.itertuples():
d=core.get(row.immutable_id)
if d is None: continue
dep={legacy(t) for t in d.get("deprecated_tasks") or []}
for k,ct in (d.get("calc_types") or {}).items():
tid=legacy(k); t=tasks.get(tid)
if t is None: continue
run,_,typ=ct.partition(" ")
if run not in ("GGA","GGA+U") or typ not in ("Static","Structure Optimization"): continue
is_lemat = tid==row.lemat_task
is_mp = tid==row.mp_task
if is_lemat or is_mp or (tid not in dep and t.get("nsites")==row.nsites and (not isinstance(row.mp_run,str) or run==row.mp_run)):
pairs.append((row.immutable_id,tid,tid in dep,is_lemat,is_mp,ct))
p=pd.DataFrame(pairs,columns=["immutable_id","task_id","task_deprecated","is_lemat_task","is_mp_task","calc_type"])
f.to_parquet(W/"identify.parquet"); p.to_parquet(W/"candidates.parquet")
print(f.review.value_counts().to_dict(), "pairs",len(p),"tasks",p.task_id.nunique(), flush=True)
return p
def extract(doc):
o=doc.get("output") or {}; s=o.get("structure") or {}
c=(doc.get("calcs_reversed") or [{}])[0]; co=c.get("output") or {}; oc=co.get("outcar") or {}
inp=doc.get("input") or {}
lat=s["lattice"]["matrix"]
return {"task_id":doc["task_id"],"energy":o.get("energy"),"lattice":lat,
"species":[x["species"][0]["element"] for x in s["sites"]],
"xyz":[x["xyz"] if "xyz" in x else list(np.dot(x["abc"],lat)) for x in s["sites"]],
"magmom":[(x.get("properties") or {}).get("magmom") for x in s["sites"]],
"forces":o.get("forces"),"stress":o.get("stress"),
"total_magnetization":oc.get("total_magnetization"),"efermi":co.get("efermi"),
"potcars":[p.get("titel") for p in (inp.get("potcar_spec") or [])],
"hubbards":inp.get("hubbards"),"last_updated":str(doc.get("last_updated")),
"calc_type":doc.get("calc_type"),"nions_steps":len(co.get("ionic_steps") or [])}
def scan(key, needed):
wanted=set(needed); got={}
for a in range(4):
try:
with requests.get(S3+key,stream=True,timeout=900) as resp:
resp.raise_for_status()
with gzip.GzipFile(fileobj=resp.raw) as fh:
for line in fh:
ids={m.decode() for m in TASK_ID_RE.findall(line[:4000])} or {m.decode() for m in TASK_ID_RE.findall(line)}
if not ids & wanted: continue
d=json.loads(line); t=d.get("task_id")
if t in wanted:
try: got[t]=extract(d)
except Exception as e: got[t]={"task_id":t,"error":repr(e)}
wanted.discard(t)
if not wanted: break
return key, got, sorted(wanted), True
except Exception as e:
time.sleep(5*(a+1))
return key, got, sorted(wanted), False
def fetch(p, workers):
need=set(p.task_id)
done=set()
for fn in OUT.glob("*.jsonl"):
for line in open(fn): done.add(json.loads(line)["task_id"])
need-=done
a2=pd.read_parquet(MAN/"manifest_atomate2.parquet",columns=["task_id","sourced_from_path"])
a2=a2[a2.task_id.isin(need)]
leg=pd.read_parquet(MAN/"manifest_legacy.parquet",columns=["task_id","nelements","spacegroup","dt"])
leg=leg[leg.task_id.isin(need-set(a2.task_id))]
keys={}
for t,k in zip(a2.task_id,a2.sourced_from_path.str.replace("s3://materialsproject-parsed/","",regex=False)): keys.setdefault(k,[]).append(t)
for r in leg.itertuples(): keys.setdefault(f"tasks-legacy/tasks/nelements={r.nelements}/output_spacegroup_number={r.spacegroup}/dt={r.dt}.jsonl.gz",[]).append(r.task_id)
nowhere=need-set(a2.task_id)-set(leg.task_id)
print("to fetch",len(need),"files",len(keys),"in no manifest",len(nowhere),flush=True)
out=open(OUT/f"part_{int(time.time())}.jsonl","w"); miss=open(OUT/"missing.txt","a")
with ProcessPoolExecutor(workers) as ex:
futs=[ex.submit(scan,k,v) for k,v in keys.items()]
for i,fu in enumerate(as_completed(futs),1):
k,got,left,ok=fu.result()
for d in got.values(): out.write(json.dumps(d)+"\n")
for t in left: miss.write(f"{t}\t{k}\t{ok}\n")
if i%100==0 or i==len(futs): out.flush(); print(f"files {i}/{len(futs)}",flush=True)
for t in nowhere: miss.write(f"{t}\tNO_MANIFEST\tTrue\n")
out.close(); miss.close()
if __name__=="__main__":
p=select() if not (W/"candidates.parquet").exists() else pd.read_parquet(W/"candidates.parquet")
fetch(p, int(sys.argv[2]) if len(sys.argv)>2 else 12)
print("DONE",flush=True)
|