File size: 3,985 Bytes
aa0469e | 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 | """Minimal loader for the solve-anywhere linear-system dataset (numpy + scipy only).
from load import catalog, runs, systems, load
for run in runs(problem="ex16p", max_n=1e6): # filter the catalog
for rec in systems(run, "heat"): # index records of one stream
A, b, x = load(run, "heat", rec["sys"]) # scipy CSR, numpy arrays (x may be None)
Command line:
python load.py summary of the catalog
python load.py <run> <stream> <i> print one system's metadata and shapes
python load.py --mtx <run> <stream> <i> <out_prefix> write <out_prefix>.mtx / _b.mtx (Matrix Market)
"""
import json
import os
import sys
import numpy as np
import scipy.io
import scipy.sparse as sp
ROOT = os.path.dirname(os.path.abspath(__file__))
def catalog():
"""Latest catalog entry per run (a re-run replaces an earlier entry)."""
latest = {}
with open(os.path.join(ROOT, "catalog.jsonl")) as f:
for line in f:
c = json.loads(line)
latest[c["path"]] = c
return list(latest.values())
def runs(source=None, problem=None, min_n=0, max_n=float("inf"), ok_only=True):
"""Run directories (absolute) matching the filters."""
out = []
for c in catalog():
if source and c["source"] != source or problem and c["problem"] != problem:
continue
if ok_only and c["exit"] != 0:
continue
ns = [n for s in c["streams"].values() for n in s["n"]]
if ns and min_n <= max(ns) <= max_n:
out.append(os.path.join(ROOT, c["path"]))
return out
def meta(run):
with open(os.path.join(run, "meta.json")) as f:
return json.load(f)
def systems(run, stream):
"""Per-system metadata (one dict per saved system, in order)."""
with open(os.path.join(run, stream, "index.jsonl")) as f:
return [json.loads(line) for line in f]
def load(run, stream, i, with_extra=False):
"""(A, b, x) of system i; x is None if not saved. with_extra=True also returns
a dict of the other saved vectors (x0, contact_dofs, block_id, ...)."""
rec = systems(run, stream)[i]
A = sp.load_npz(os.path.join(run, stream, f"A_{rec['A']:04d}.npz"))
v = np.load(os.path.join(run, stream, f"sys_{i:05d}.npz"))
x = v["x"] if "x" in v.files else None
if with_extra:
return A, v["b"], x, {k: v[k] for k in v.files if k not in ("b", "x")}
return A, v["b"], x
def to_matrix_market(run, stream, i, prefix):
A, b, x = load(run, stream, i)
scipy.io.mmwrite(prefix + ".mtx", A, precision=17)
scipy.io.mmwrite(prefix + "_b.mtx", b.reshape(-1, 1), precision=17)
if x is not None:
scipy.io.mmwrite(prefix + "_x.mtx", x.reshape(-1, 1), precision=17)
if __name__ == "__main__":
if len(sys.argv) >= 6 and sys.argv[1] == "--mtx":
to_matrix_market(os.path.abspath(sys.argv[2]), sys.argv[3], int(sys.argv[4]), sys.argv[5])
elif len(sys.argv) == 4:
run, stream, i = os.path.abspath(sys.argv[1]), sys.argv[2], int(sys.argv[3])
A, b, x, extra = load(run, stream, i, with_extra=True)
print(json.dumps(systems(run, stream)[i], indent=2))
print("A", A.shape, A.nnz, "| b", b.shape, "| x", None if x is None else x.shape, "| extra", list(extra))
else:
by = {}
for c in catalog():
k = (c["source"], c["problem"])
nsys = sum(s["systems"] for s in c["streams"].values())
nmax = max([n for s in c["streams"].values() for n in s["n"]] or [0])
r, s_, b_, m, f = by.get(k, (0, 0, 0, 0, 0))
by[k] = (r + 1, s_ + nsys, b_ + c["bytes"], max(m, nmax), f + (c["exit"] != 0))
print(f"{'source':8s} {'problem':18s} {'runs':>5s} {'failed':>6s} {'systems':>8s} {'GB':>7s} {'largest n':>10s}")
for (src, prob), (r, s_, b_, m, f) in sorted(by.items()):
print(f"{src:8s} {prob:18s} {r:5d} {f:6d} {s_:8d} {b_ / 1e9:7.2f} {m:10d}")
|