FanQY commited on
Commit
c2fc1f0
·
verified ·
1 Parent(s): df345c1

Upload folder using huggingface_hub

Browse files
tools/audit_integer_literals.py CHANGED
@@ -11,8 +11,8 @@ Reports the counterexamples rather than a verdict. A property that holds on four
11
  million programs is worth stating precisely; one that holds on 3,999,998 is worth
12
  stating precisely too, and the difference only shows up if the failures are kept.
13
 
14
- python audit_integer_literals.py --parquet-root /beegfs/scratch/qinyuan_fan/GeomCAD/data \\
15
- --out /beegfs/scratch/qinyuan_fan/results_integer_audit.json
16
  """
17
 
18
  from __future__ import annotations
 
11
  million programs is worth stating precisely; one that holds on 3,999,998 is worth
12
  stating precisely too, and the difference only shows up if the failures are kept.
13
 
14
+ python audit_integer_literals.py --parquet-root ./data \\
15
+ --out ./integer_audit.json
16
  """
17
 
18
  from __future__ import annotations
tools/verify_mesh_roundtrip.py CHANGED
@@ -12,7 +12,7 @@ the question directly.
12
  Compared on vertex and face counts, bounding box, and volume. Vertex order is
13
  not compared because tessellation is not required to enumerate in a fixed order.
14
 
15
- python verify_mesh_roundtrip.py --n 100
16
  """
17
 
18
  from __future__ import annotations
@@ -24,7 +24,9 @@ from pathlib import Path
24
  import numpy as np
25
  import trimesh
26
 
27
- WS = Path("/home/qinyuan_fan/cadllm/workspace")
 
 
28
  OPS = ["extrude", "revolve", "sweep", "loft"]
29
  # The corpus meshes come from cq.exporters.export at its defaults, not from
30
  # tessellate(0.001, 0.1). infer_cadrille_geomcad.py calls the latter and its
@@ -71,12 +73,14 @@ def build(code: str):
71
  def main() -> None:
72
  ap = argparse.ArgumentParser()
73
  ap.add_argument("--n", type=int, default=100, help="samples per operation")
 
 
74
  args = ap.parse_args()
75
 
76
  print(f"{'op':9s} {'pairs':>6s} {'rebuilt':>8s} {'exact F':>10s} "
77
  f"{'bbox<1e-6':>10s} {'vol<0.01%':>10s}")
78
  for op in OPS:
79
- root = WS / "testsets_indexed" / op
80
  pairs = rebuilt = same_vf = same_bb = same_vol = 0
81
  for i in range(args.n):
82
  py, stl = root / str(i) / f"{i}.py", root / str(i) / f"{i}.stl"
 
12
  Compared on vertex and face counts, bounding box, and volume. Vertex order is
13
  not compared because tessellation is not required to enumerate in a fixed order.
14
 
15
+ python tools/verify_mesh_roundtrip.py --n 100
16
  """
17
 
18
  from __future__ import annotations
 
24
  import numpy as np
25
  import trimesh
26
 
27
+ # Default layout is the release itself: test/<op>/<id>/<id>.py and <id>.stl,
28
+ # relative to the repository root. Override with --root.
29
+ DEFAULT_ROOT = Path(__file__).resolve().parent.parent / "test"
30
  OPS = ["extrude", "revolve", "sweep", "loft"]
31
  # The corpus meshes come from cq.exporters.export at its defaults, not from
32
  # tessellate(0.001, 0.1). infer_cadrille_geomcad.py calls the latter and its
 
73
  def main() -> None:
74
  ap = argparse.ArgumentParser()
75
  ap.add_argument("--n", type=int, default=100, help="samples per operation")
76
+ ap.add_argument("--root", type=Path, default=DEFAULT_ROOT,
77
+ help="directory holding <op>/<id>/<id>.py and <id>.stl")
78
  args = ap.parse_args()
79
 
80
  print(f"{'op':9s} {'pairs':>6s} {'rebuilt':>8s} {'exact F':>10s} "
81
  f"{'bbox<1e-6':>10s} {'vol<0.01%':>10s}")
82
  for op in OPS:
83
+ root = args.root / op
84
  pairs = rebuilt = same_vf = same_bb = same_vol = 0
85
  for i in range(args.n):
86
  py, stl = root / str(i) / f"{i}.py", root / str(i) / f"{i}.stl"