ONNX
onnxruntime
onnx-mlir
quantization
fp32
File size: 1,232 Bytes
ed3aeeb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from __future__ import annotations

import json

import numpy as np
import pytest

from scripts.stages.evaluate_sp01_mlcommons_kws import (
    canonical_digest,
    load_checkpoint,
    official_int8_input,
    quality_decision,
)


def test_official_int8_input_matches_scale_offset_cast_without_local_ptq() -> None:
    feature = np.asarray([-2.0, 0.0, 1.0, 2.0], dtype=np.float32)
    actual = official_int8_input(feature, 0.5, 3)
    assert actual.dtype == np.int8
    assert actual.tolist() == [-1, 3, 5, 7]


def test_quality_decision_uses_inclusive_mlcommons_threshold() -> None:
    assert quality_decision(9, 10, 0.9)["threshold_met"] is True
    assert quality_decision(899, 1000, 0.9)["threshold_met"] is False


def test_checkpoint_requires_exact_fingerprint(tmp_path) -> None:
    path = tmp_path / "checkpoint.jsonl"
    path.write_text(json.dumps({"fingerprint": "one", "index": 0, "label_id": 2}) + "\n")
    assert load_checkpoint(path, "one")[0]["label_id"] == 2
    with pytest.raises(ValueError, match="fingerprint mismatch"):
        load_checkpoint(path, "two")


def test_canonical_digest_is_key_order_independent() -> None:
    assert canonical_digest({"a": 1, "b": 2}) == canonical_digest({"b": 2, "a": 1})