File size: 2,812 Bytes
dfb775d | 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 | """Classroom before/after evaluation + autotune feedback loop."""
from __future__ import annotations
from mindxtrain.autotune import feedback as FB
from mindxtrain.governance.classroom import evaluate_classroom
def test_classroom_passes_when_after_recalls_persona():
inquiries = ["who are you?", "what do you do?"]
baseline = ["i am codephreak, augmentic intelligence.", "i orchestrate autonomous agents."]
before = ["I am an AI assistant.", "I help with tasks."]
after = baseline # perfect recall after imprint
rep = evaluate_classroom(inquiries, before, after, baseline)
assert rep.recall > rep.before_recall
assert rep.imprint_delta > 0
assert rep.persona_maintained is True
assert rep.pairwise_after_better == 1.0
assert rep.passed is True
def test_classroom_fails_with_no_movement():
inquiries = ["who?"]
baseline = ["i am codephreak."]
same = ["I am an AI."]
rep = evaluate_classroom(inquiries, same, same, baseline)
assert rep.passed is False
assert rep.persona_maintained is False
def test_feedback_record_and_read(tmp_path):
p = tmp_path / "feedback.jsonl"
FB.record(run_id="r1", params={"epochs": 12, "grad_accum": 1, "per_device": 1},
classroom_score=0.04, passed=True, boardroom_outcome="approved",
timestamp="2026-06-12T00:00:00+00:00", path=p)
rows = FB.read_all(p)
assert len(rows) == 1
assert rows[0].run_id == "r1" and rows[0].boardroom_outcome == "approved"
def test_suggest_next_params_trains_harder_on_failure():
nxt = FB.suggest_next_params({"epochs": 12, "grad_accum": 4, "per_device": 1},
passed=False, classroom_score=0.0)
assert nxt["epochs"] > 12 # more epochs
assert nxt["grad_accum"] == 1 # forced to 1
def test_suggest_next_params_nudges_weak_imprint():
nxt = FB.suggest_next_params({"epochs": 12, "grad_accum": 1, "per_device": 1},
passed=True, classroom_score=0.02)
assert nxt["epochs"] == 16
def test_suggest_next_params_keeps_on_good_pass():
base = {"epochs": 12, "grad_accum": 1, "per_device": 1}
assert FB.suggest_next_params(base, passed=True, classroom_score=0.4) == base
def test_suggest_from_history(tmp_path):
p = tmp_path / "f.jsonl"
assert FB.suggest_from_history({"epochs": 10, "grad_accum": 2, "per_device": 1}, path=p) == \
{"epochs": 10, "grad_accum": 2, "per_device": 1} # empty → defaults
FB.record(run_id="r", params={"epochs": 12, "grad_accum": 2, "per_device": 1},
classroom_score=0.0, passed=False, path=p)
nxt = FB.suggest_from_history({"epochs": 10, "grad_accum": 2, "per_device": 1}, path=p)
assert nxt["epochs"] > 12 and nxt["grad_accum"] == 1 # nudged from the failed run
|