| 1 | # SPDX-License-Identifier: Apache-2.0 |
| 2 | # SPDX-FileCopyrightText: Copyright contributors to the vLLM project |
| 3 | """Tests whether unsupported GPTQ group actorder models are rejected. |
| 4 | |
| 5 | Run `pytest tests/quantization/test_gptq_dynamic.py --forked`. |
| 6 | |
| 7 | Note: Only symmetric GPTQ models are supported after consolidation to Marlin. |
| 8 | """ |
| 9 | |
| 10 | import pytest |
| 11 | |
| 12 | MODELS = [ |
| 13 | "ModelCloud/Qwen1.5-1.8B-Chat-GPTQ-4bits-dynamic-cfg-with-lm_head-symTrue", |
| 14 | ] |
| 15 | |
| 16 | |
| 17 | @pytest.mark.parametrize("model_id", MODELS) |
| 18 | def test_gptq_with_dynamic_desc_act_rejected(vllm_runner, model_id: str): |
| 19 | with ( |
| 20 | pytest.raises(ValueError, match="desc_act=True"), |
| 21 | vllm_runner(model_id, max_model_len=2048, enforce_eager=True), |
| 22 | ): |
| 23 | pass |