| """Device-resolution regressions for local and GPU execution paths.""" | |
| import pytest | |
| import torch | |
| from aegis.configs.dataclasses import EvidenceConfig | |
| from aegis.models.nli_model import NliModel | |
| from aegis.utils.device import get_device | |
| def test_cpu_device_is_always_explicitly_resolvable(): | |
| assert get_device("cpu") == torch.device("cpu") | |
| def test_nli_model_honors_evidence_device_config(): | |
| model = NliModel(EvidenceConfig(device="cpu", is_mock=True)) | |
| assert model.device == torch.device("cpu") | |
| def test_cuda_request_fails_clearly_when_unavailable(): | |
| if torch.cuda.is_available(): | |
| pytest.skip("CUDA is available; the unavailable-device guard is not applicable") | |
| with pytest.raises(RuntimeError, match="CUDA was requested but is unavailable"): | |
| get_device("cuda") | |
Xet Storage Details
- Size:
- 807 Bytes
- Xet hash:
- a995b099c1a5b8043c1ee88a8c2c4aec937c15cea591ca672031db44fd945bda
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.