PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
Fix: temporarily skip tests due to known issues with xfail annotations (#2113)
chen, suyue60b813cb3bad
diff --git a/test/test_cpu/export/test_llmc_format.py b/test/test_cpu/export/test_llmc_format.pyindex 0a3c3282a..6c2dd01b6 100644--- a/test/test_cpu/export/test_llmc_format.py+++ b/test/test_cpu/export/test_llmc_format.py@@ -267,8 +267,6 @@ def test_llmcompressor_static_fp8_kv_config(tiny_opt_model_path, dataloader, tmpassert kv_cache_scheme["symmetric"] is True-# TODO: remove xfail once the issue is resolved.[email protected](reason="https://github.com/vllm-project/compressed-tensors/issues/795")def test_llmcompressor_static_fp8_attention_config(dataloader, tmp_path):model_name = get_model_path("stas/tiny-random-llama-2")autoround = AutoRound(diff --git a/test/test_xpu/test_llmc_integration.py b/test/test_xpu/test_llmc_integration.pyindex 3e68bd144..cb0471a65 100644--- a/test/test_xpu/test_llmc_integration.py+++ b/test/test_xpu/test_llmc_integration.py@@ -74,6 +74,8 @@ w8a8_static_recipe_modifier = AutoRoundModifier()+# TODO: remove xfail once the issue is resolved.[email protected](reason="skip this case temporarily due to issue https://github.com/intel/auto-round/issues/2112")@pytest.mark.skipif(torch.xpu.device_count() < 1, reason="test requires at least 1 XPU")@pytest.mark.parametrize("recipe",