File size: 1,621 Bytes
1ba6b4e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1731700
1ba6b4e
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
import io

from PIL import Image

import runner


class FakePipeline:
    def __init__(self):
        self.calls = []

    def __call__(self, **kwargs):
        self.calls.append(kwargs)
        seed = kwargs["generator"].initial_seed()
        return type("Output", (), {"images": [Image.new("RGB", (kwargs["width"], kwargs["height"]), (seed % 255, 40, 90))]})()


class SafeClassifier:
    def __call__(self, image):
        return [{"label": "normal", "score": 0.999}, {"label": "nsfw", "score": 0.001}]


def test_grid_uses_coordinator_controls_and_distinct_seeds(monkeypatch):
    pipe = FakePipeline()
    monkeypatch.setattr(runner, "PIPELINE", pipe)
    monkeypatch.setattr(runner, "SAFETY", SafeClassifier())
    task = {"id": "j", "prompt": "rat", "operation": "grid", "seed": 10,
            "outputCount": 4, "width": 512, "height": 512, "steps": 16,
            "assignmentToken": "a"}
    result = runner.render_task(task, None)
    assert len(result["images"]) == 4
    assert len({call["generator"].initial_seed() for call in pipe.calls}) == 4
    assert all(call["num_inference_steps"] == 16 for call in pipe.calls)
    grid = Image.open(io.BytesIO(result["grid"]["bytes"]))
    assert grid.size == (1024, 1024)
    assert result["grid"]["localSafety"]["unsafe"] is False


def test_gpu_duration_is_bounded():
    assert runner.gpu_duration({"operation": "grid", "outputCount": 4, "steps": 16}) > runner.gpu_duration({"outputCount": 1, "steps": 16})
    assert runner.gpu_duration({"outputCount": 1, "steps": 16}, b"source") >= 60
    assert runner.gpu_duration({"outputCount": 99, "steps": 999}) <= 300