PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
Add envs.py for environment checks and update imports in test files (#1432)
Xin He388e6c0b364d
diff --git a/auto_round/testing_utils.py b/test/envs.pysimilarity index 100%rename from auto_round/testing_utils.pyrename to test/envs.pydiff --git a/test/test_cpu/backends/test_torch_backend.py b/test/test_cpu/backends/test_torch_backend.pyindex 814dc08ee..60f7c0662 100644--- a/test/test_cpu/backends/test_torch_backend.py+++ b/test/test_cpu/backends/test_torch_backend.py@@ -5,8 +5,8 @@ import torchfrom transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import require_autogptq, require_gptqmodel+from ...envs import require_autogptq, require_gptqmodelfrom ...helpers import evaluate_accuracy, get_model_path, model_inferdiff --git a/test/test_cpu/quantization/test_mix_bits.py b/test/test_cpu/quantization/test_mix_bits.pyindex 8affa527a..062afb058 100644--- a/test/test_cpu/quantization/test_mix_bits.py+++ b/test/test_cpu/quantization/test_mix_bits.py@@ -8,8 +8,8 @@ import torchfrom transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import require_gptqmodel+from ...envs import require_gptqmodelfrom ...helpers import evaluate_accuracy, opt_name_or_pathdiff --git a/test/test_cpu/quantization/test_mxfp_save_load.py b/test/test_cpu/quantization/test_mxfp_save_load.pyindex 396c47735..5e12edc68 100644--- a/test/test_cpu/quantization/test_mxfp_save_load.py+++ b/test/test_cpu/quantization/test_mxfp_save_load.py@@ -12,8 +12,8 @@ from auto_round.experimental import qmodules as ar_qmodulesfrom auto_round.export.export_to_autoround import qlinear_fp as ar_qlinear_fpfrom auto_round.formats import AutoRoundExportFormatfrom auto_round.inference.backend import MX_TENSOR_DATA_TYPES-from auto_round.testing_utils import has_module+from ...envs import has_modulefrom ...helpers import get_model_pathtesting_scheme_name_lst = [diff --git a/test/test_cuda/advanced/test_multiple_card.py b/test/test_cuda/advanced/test_multiple_card.pyindex 33ac5a38d..08f2b4676 100644--- a/test/test_cuda/advanced/test_multiple_card.py+++ b/test/test_cuda/advanced/test_multiple_card.py@@ -6,8 +6,8 @@ from lm_eval.utils import make_table # pylint: disable=E0401from transformers import AutoModelForCausalLM, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import multi_card, require_gptqmodel, require_greater_than_050+from ...envs import multi_card, require_gptqmodel, require_greater_than_050from ...helpers import evaluate_accuracy, get_model_path, get_tiny_modeldiff --git a/test/test_cuda/advanced/test_multiple_card_calib.py b/test/test_cuda/advanced/test_multiple_card_calib.pyindex 7b4f4d6ea..e41e0cfd7 100644--- a/test/test_cuda/advanced/test_multiple_card_calib.py+++ b/test/test_cuda/advanced/test_multiple_card_calib.py@@ -5,7 +5,7 @@ import sysimport pytest-from auto_round.testing_utils import multi_card+from ...envs import multi_cardAUTO_ROUND_PATH = __file__.split("/")AUTO_ROUND_PATH = "/".join(AUTO_ROUND_PATH[: AUTO_ROUND_PATH.index("test")])diff --git a/test/test_cuda/backends/test_exllamav2_backend.py b/test/test_cuda/backends/test_exllamav2_backend.pyindex 07343ca2a..618917df8 100644--- a/test/test_cuda/backends/test_exllamav2_backend.py+++ b/test/test_cuda/backends/test_exllamav2_backend.py@@ -7,8 +7,8 @@ from transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizerfrom auto_round import (AutoRound,)-from auto_round.testing_utils import require_autogptq, require_gptqmodel, require_package_version_ut+from ...envs import require_autogptq, require_gptqmodel, require_package_version_utfrom ...helpers import evaluate_accuracy, get_model_path, model_inferdiff --git a/test/test_cuda/backends/test_torch_backend.py b/test/test_cuda/backends/test_torch_backend.pyindex 87aed9073..d175f80a7 100644--- a/test/test_cuda/backends/test_torch_backend.py+++ b/test/test_cuda/backends/test_torch_backend.py@@ -5,8 +5,8 @@ import torchfrom transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import require_autogptq, require_gptqmodel+from ...envs import require_autogptq, require_gptqmodelfrom ...helpers import evaluate_accuracy, get_model_path, model_inferdiff --git a/test/test_cuda/backends/test_triton_backend.py b/test/test_cuda/backends/test_triton_backend.pyindex 545ed013f..5200494f5 100644--- a/test/test_cuda/backends/test_triton_backend.py+++ b/test/test_cuda/backends/test_triton_backend.py@@ -5,8 +5,8 @@ import torchfrom transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import require_greater_than_050+from ...envs import require_greater_than_050from ...helpers import evaluate_accuracy, model_inferdiff --git a/test/test_cuda/core/test_main_func.py b/test/test_cuda/core/test_main_func.pyindex 82ccbea44..f39458451 100644--- a/test/test_cuda/core/test_main_func.py+++ b/test/test_cuda/core/test_main_func.py@@ -9,8 +9,8 @@ from transformers import AutoModelForCausalLM, AutoTokenizerfrom transformers.utils.versions import require_versionfrom auto_round import AutoRound, AutoRoundAdam-from auto_round.testing_utils import require_awq, require_gptqmodel, require_optimum, require_package_version_ut+from ...envs import require_awq, require_gptqmodel, require_optimum, require_package_version_utfrom ...helpers import evaluate_accuracy, get_model_pathdiff --git a/test/test_cuda/export/test_auto_round_format.py b/test/test_cuda/export/test_auto_round_format.pyindex 3eff1119b..e5600a98b 100644--- a/test/test_cuda/export/test_auto_round_format.py+++ b/test/test_cuda/export/test_auto_round_format.py@@ -7,14 +7,14 @@ import transformersfrom transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import (++from ...envs import (require_autogptq,require_awq,require_greater_than_050,require_ipex,require_package_version_ut,)-from ...helpers import evaluate_accuracy, get_model_path, get_tiny_model, model_inferdiff --git a/test/test_cuda/export/test_export.py b/test/test_cuda/export/test_export.pyindex 4eb582c32..f5e2e1af0 100644--- a/test/test_cuda/export/test_export.py+++ b/test/test_cuda/export/test_export.py@@ -8,8 +8,8 @@ from packaging import versionfrom transformers import AutoConfig, AutoModelForCausalLM, AutoRoundConfig, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import require_awq, require_optimum, require_package_version_ut+from ...envs import require_awq, require_optimum, require_package_version_utfrom ...helpers import get_model_path, get_tiny_model, transformers_versiondiff --git a/test/test_cuda/export/test_gguf.py b/test/test_cuda/export/test_gguf.pyindex 11159844b..e81a0e5b0 100644--- a/test/test_cuda/export/test_gguf.py+++ b/test/test_cuda/export/test_gguf.py@@ -9,8 +9,8 @@ from packaging import versionfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import require_gguf+from ...envs import require_gguffrom ...helpers import evaluate_accuracy, get_model_path, get_tiny_model, save_tiny_model, transformers_versionAUTO_ROUND_PATH = __file__.split("/")diff --git a/test/test_cuda/integrations/test_transformers.py b/test/test_cuda/integrations/test_transformers.pyindex b99a3678f..7adbe43d7 100644--- a/test/test_cuda/integrations/test_transformers.py+++ b/test/test_cuda/integrations/test_transformers.py@@ -26,8 +26,7 @@ from transformers.testing_utils import ()from transformers.utils import is_torch_available-from auto_round.testing_utils import is_ipex_available-+from ...envs import is_ipex_availablefrom ...helpers import get_model_pathif is_torch_available():diff --git a/test/test_cuda/models/test_conv1d.py b/test/test_cuda/models/test_conv1d.pyindex 8c1f654a3..deae349b3 100644--- a/test/test_cuda/models/test_conv1d.py+++ b/test/test_cuda/models/test_conv1d.py@@ -6,8 +6,8 @@ import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import require_gptqmodel+from ...envs import require_gptqmodelfrom ...helpers import get_model_path, get_tiny_model, model_inferdiff --git a/test/test_cuda/models/test_diffusion.py b/test/test_cuda/models/test_diffusion.pyindex 80bc238e8..703653be7 100644--- a/test/test_cuda/models/test_diffusion.py+++ b/test/test_cuda/models/test_diffusion.py@@ -9,8 +9,8 @@ from packaging import versionfrom PIL import Imagefrom auto_round import AutoRoundDiffusion-from auto_round.testing_utils import require_gptqmodel, require_optimum, require_vlm_env+from ...envs import require_gptqmodel, require_optimum, require_vlm_envfrom ...helpers import transformers_versiondiff --git a/test/test_cuda/models/test_support_vlms.py b/test/test_cuda/models/test_support_vlms.pyindex 67296925a..7b4e89177 100644--- a/test/test_cuda/models/test_support_vlms.py+++ b/test/test_cuda/models/test_support_vlms.py@@ -8,8 +8,7 @@ from packaging import versionfrom PIL import Imagefrom transformers import AutoRoundConfig # # must import for auto-round format-from auto_round.testing_utils import require_gptqmodel, require_package_version_ut, require_vlm_env-+from ...envs import require_gptqmodel, require_package_version_ut, require_vlm_envfrom ...helpers import transformers_versionAUTO_ROUND_PATH = __file__.split("/")diff --git a/test/test_cuda/models/test_vlms.py b/test/test_cuda/models/test_vlms.pyindex ad7fcaa11..3cbdc2a67 100644--- a/test/test_cuda/models/test_vlms.py+++ b/test/test_cuda/models/test_vlms.py@@ -8,7 +8,7 @@ import requestsfrom PIL import Imagefrom transformers import AutoRoundConfig-from auto_round.testing_utils import require_gptqmodel, require_optimum, require_vlm_env+from ...envs import require_gptqmodel, require_optimum, require_vlm_envclass TestAutoRound:diff --git a/test/test_cuda/quantization/test_2_3bits.py b/test/test_cuda/quantization/test_2_3bits.pyindex 6caccdbce..90637d03f 100644--- a/test/test_cuda/quantization/test_2_3bits.py+++ b/test/test_cuda/quantization/test_2_3bits.py@@ -8,8 +8,8 @@ from lm_eval.utils import make_table # pylint: disable=E0401from transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import require_autogptq, require_greater_than_050, require_greater_than_051+from ...envs import require_autogptq, require_greater_than_050, require_greater_than_051from ...helpers import evaluate_accuracy, get_model_path, model_inferdiff --git a/test/test_cuda/quantization/test_mix_bits.py b/test/test_cuda/quantization/test_mix_bits.pyindex eee5874a0..5f7175173 100644--- a/test/test_cuda/quantization/test_mix_bits.py+++ b/test/test_cuda/quantization/test_mix_bits.py@@ -8,12 +8,12 @@ import torchfrom transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import (++from ...envs import (require_awq,require_gptqmodel,require_package_version_ut,)-from ...helpers import get_model_pathdiff --git a/test/test_cuda/quantization/test_mxfp_and_nvfp_quant.py b/test/test_cuda/quantization/test_mxfp_and_nvfp_quant.pyindex d76b556e3..c3f3e09c7 100644--- a/test/test_cuda/quantization/test_mxfp_and_nvfp_quant.py+++ b/test/test_cuda/quantization/test_mxfp_and_nvfp_quant.py@@ -10,8 +10,8 @@ from auto_round import schemes as ar_schemesfrom auto_round.experimental import qmodules as ar_qmodulesfrom auto_round.export.export_to_autoround import qlinear_fp as ar_qlinear_fpfrom auto_round.formats import AutoRoundExportFormat-from auto_round.testing_utils import has_module+from ...envs import has_modulefrom ...helpers import get_model_pathtesting_schemes = [diff --git a/test/test_cuda/quantization/test_mxfp_nvfp.py b/test/test_cuda/quantization/test_mxfp_nvfp.pyindex 75fe039ce..b9263f5d9 100644--- a/test/test_cuda/quantization/test_mxfp_nvfp.py+++ b/test/test_cuda/quantization/test_mxfp_nvfp.py@@ -7,8 +7,8 @@ import transformersfrom transformers import AutoConfig, AutoModelForCausalLM, AutoTokenizerfrom auto_round import AutoRound-from auto_round.testing_utils import require_awq, require_optimum+from ...envs import require_awq, require_optimumfrom ...helpers import get_model_path, save_tiny_modeldiff --git a/test/test_cuda/schemes/test_auto_scheme.py b/test/test_cuda/schemes/test_auto_scheme.pyindex ecfc7cb34..6d23b3de1 100644--- a/test/test_cuda/schemes/test_auto_scheme.py+++ b/test/test_cuda/schemes/test_auto_scheme.py@@ -8,9 +8,9 @@ from transformers import AutoRoundConfigfrom auto_round import AutoRound, AutoSchemefrom auto_round.auto_scheme.utils import compute_avg_bits_for_model-from auto_round.testing_utils import multi_cardfrom auto_round.utils import get_module+from ...envs import multi_cardfrom ...helpers import evaluate_accuracy, get_model_path, get_tiny_model