Buckets:

download
raw
1.82 kB
import os
from pathlib import Path
import yaml
from nancy_brain.weights_persistence import load_model_weights, save_model_weights, set_model_weight
def test_load_missing_returns_empty(tmp_path):
p = tmp_path / "model_weights.yaml"
assert not p.exists()
data = load_model_weights(p)
assert isinstance(data, dict)
assert data == {}
def test_save_and_load(tmp_path):
p = tmp_path / "model_weights.yaml"
mapping = {"doc/a": 1, "doc/b": "2"}
save_model_weights(mapping, p)
assert p.exists()
loaded = load_model_weights(p)
# Values should be coerced to floats
assert loaded["doc/a"] == 1.0
assert loaded["doc/b"] == 2.0
def test_set_model_weight_add_update_remove(tmp_path):
p = tmp_path / "model_weights.yaml"
# Add
prev = set_model_weight(p, "x/1", 1.2)
assert prev is None
loaded = load_model_weights(p)
assert loaded.get("x/1") == 1.2
# Update
prev2 = set_model_weight(p, "x/1", 1.5)
assert float(prev2) == 1.2
loaded = load_model_weights(p)
assert loaded.get("x/1") == 1.5
# Remove
prev3 = set_model_weight(p, "x/1", None)
assert float(prev3) == 1.5
loaded = load_model_weights(p)
assert "x/1" not in loaded
def test_load_invalid_content(tmp_path):
p = tmp_path / "model_weights.yaml"
# Write a YAML list (invalid for our loader)
p.write_text(yaml.dump([1, 2, 3]))
loaded = load_model_weights(p)
assert loaded == {}
def test_coerce_invalid_value_skips(tmp_path):
p = tmp_path / "model_weights.yaml"
# Write mapping with a non-numeric value
p.write_text(yaml.dump({"good": 2, "bad": "nope"}))
loaded = load_model_weights(p)
assert "good" in loaded and loaded["good"] == 2.0
# 'bad' should be skipped during coercion
assert "bad" not in loaded

Xet Storage Details

Size:
1.82 kB
·
Xet hash:
b3a439ad3f0c43614736fff91e1e945d417c75a6ead0ea52e7799496623fe0d2

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.