| 1 | # SPDX-License-Identifier: Apache-2.0 |
| 2 | # SPDX-FileCopyrightText: Copyright contributors to the vLLM project |
| 3 | import contextlib |
| 4 | import io |
| 5 | from types import SimpleNamespace |
| 6 | from unittest.mock import Mock |
| 7 | |
| 8 | import pytest |
| 9 | import torch |
| 10 | |
| 11 | from vllm.model_executor.model_loader import tensorizer |
| 12 | |
| 13 | pytestmark = [pytest.mark.cpu_test, pytest.mark.skip_global_cleanup] |
| 14 | |
| 15 | |
| 16 | def test_serialize_vllm_model_forwards_model_revision( |
| 17 | monkeypatch: pytest.MonkeyPatch, |
| 18 | ): |
| 19 | @contextlib.contextmanager |
| 20 | def fake_open_stream(*args, **kwargs): |
| 21 | yield io.BytesIO() |
| 22 | |
| 23 | snapshot_download = Mock() |
| 24 | monkeypatch.setattr(tensorizer, "open_stream", fake_open_stream) |
| 25 | monkeypatch.setattr(tensorizer, "TensorSerializer", Mock()) |
| 26 | monkeypatch.setattr( |
| 27 | tensorizer, |
| 28 | "hf_api", |
| 29 | lambda: SimpleNamespace(snapshot_download=snapshot_download), |
| 30 | ) |
| 31 | |
| 32 | model = torch.nn.Module() |
| 33 | tensorizer_config = SimpleNamespace( |
| 34 | _construct_tensorizer_args=lambda: SimpleNamespace( |
| 35 | tensorizer_uri="/tmp/model.tensors", |
| 36 | tensorizer_dir="/tmp/out", |
| 37 | stream_kwargs={}, |
| 38 | ), |
| 39 | encryption_keyfile=None, |
| 40 | _is_sharded=False, |
| 41 | serialization_kwargs=None, |
| 42 | ) |
| 43 | model_config = SimpleNamespace( |
| 44 | served_model_name="repo/model", |
| 45 | revision="pinned-revision", |
| 46 | ) |
| 47 | |
| 48 | tensorizer.serialize_vllm_model(model, tensorizer_config, model_config) |
| 49 | |
| 50 | assert snapshot_download.call_args.args == ("repo/model",) |
| 51 | assert snapshot_download.call_args.kwargs["revision"] == "pinned-revision" |