DeepSeek-OCR-2-Math / tests /test_spatial_blocks.py
ricklon's picture
Refine equation grounding with zoom-in pass and per-box spatial blocks
f76cb58
import ast
import pathlib
import re
import unittest
def _load_grounding_blocks():
app_path = pathlib.Path(__file__).resolve().parents[1] / "app.py"
source = app_path.read_text(encoding="utf-8")
module = ast.parse(source, filename=str(app_path))
wanted = {
"_parse_coord_payload",
"_extract_grounding_entries",
"_grounding_blocks_from_raw",
}
fn_nodes = [n for n in module.body if isinstance(n, ast.FunctionDef) and n.name in wanted]
fn_nodes.sort(key=lambda n: n.lineno)
test_mod = ast.Module(body=fn_nodes, type_ignores=[])
code = compile(test_mod, filename=str(app_path), mode="exec")
scope = {
"ast": ast,
"re": re,
"GROUNDING_PATTERN": re.compile(r"<\|ref\|>(.*?)<\|/ref\|><\|det\|>(.*?)<\|/det\|>", re.DOTALL),
}
exec(code, scope)
return scope["_grounding_blocks_from_raw"]
class SpatialBlockTests(unittest.TestCase):
def test_multi_coord_refs_render_as_multiple_blocks(self):
grounding_blocks = _load_grounding_blocks()
raw = (
"Equation cluster\n"
"<|ref|>formula<|/ref|><|det|>[[100,100,600,220],[100,240,600,360]]<|/det|>\n"
"Trailing text\n"
"<|ref|>text<|/ref|><|det|>[[40,400,700,520]]<|/det|>"
)
blocks = grounding_blocks(raw)
self.assertEqual(3, len(blocks))
formula_blocks = [b for b in blocks if b["label"] == "formula"]
self.assertEqual(2, len(formula_blocks))
self.assertEqual(100.0, formula_blocks[0]["x1"])
self.assertEqual(600.0, formula_blocks[0]["x2"])
self.assertEqual(100.0, formula_blocks[1]["x1"])
self.assertEqual(600.0, formula_blocks[1]["x2"])
self.assertEqual("Equation cluster", formula_blocks[0]["text"])
self.assertEqual("", formula_blocks[1]["text"])
if __name__ == "__main__":
unittest.main()