Skip to content

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.py
similarity index 100%
rename from auto_round/testing_utils.py
rename to test/envs.py
diff --git a/test/test_cpu/backends/test_torch_backend.py b/test/test_cpu/backends/test_torch_backend.py
index 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 torch
from transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizer
from auto_round import AutoRound
-from auto_round.testing_utils import require_autogptq, require_gptqmodel
+from ...envs import require_autogptq, require_gptqmodel
from ...helpers import evaluate_accuracy, get_model_path, model_infer
diff --git a/test/test_cpu/quantization/test_mix_bits.py b/test/test_cpu/quantization/test_mix_bits.py
index 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 torch
from transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizer
from auto_round import AutoRound
-from auto_round.testing_utils import require_gptqmodel
+from ...envs import require_gptqmodel
from ...helpers import evaluate_accuracy, opt_name_or_path
diff --git a/test/test_cpu/quantization/test_mxfp_save_load.py b/test/test_cpu/quantization/test_mxfp_save_load.py
index 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_qmodules
from auto_round.export.export_to_autoround import qlinear_fp as ar_qlinear_fp
from auto_round.formats import AutoRoundExportFormat
from auto_round.inference.backend import MX_TENSOR_DATA_TYPES
-from auto_round.testing_utils import has_module
+from ...envs import has_module
from ...helpers import get_model_path
testing_scheme_name_lst = [
diff --git a/test/test_cuda/advanced/test_multiple_card.py b/test/test_cuda/advanced/test_multiple_card.py
index 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=E0401
from transformers import AutoModelForCausalLM, AutoTokenizer
from 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_050
from ...helpers import evaluate_accuracy, get_model_path, get_tiny_model
diff --git a/test/test_cuda/advanced/test_multiple_card_calib.py b/test/test_cuda/advanced/test_multiple_card_calib.py
index 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 sys
import pytest
-from auto_round.testing_utils import multi_card
+from ...envs import multi_card
AUTO_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.py
index 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, AutoTokenizer
from 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_ut
from ...helpers import evaluate_accuracy, get_model_path, model_infer
diff --git a/test/test_cuda/backends/test_torch_backend.py b/test/test_cuda/backends/test_torch_backend.py
index 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 torch
from transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizer
from auto_round import AutoRound
-from auto_round.testing_utils import require_autogptq, require_gptqmodel
+from ...envs import require_autogptq, require_gptqmodel
from ...helpers import evaluate_accuracy, get_model_path, model_infer
diff --git a/test/test_cuda/backends/test_triton_backend.py b/test/test_cuda/backends/test_triton_backend.py
index 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 torch
from transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizer
from auto_round import AutoRound
-from auto_round.testing_utils import require_greater_than_050
+from ...envs import require_greater_than_050
from ...helpers import evaluate_accuracy, model_infer
diff --git a/test/test_cuda/core/test_main_func.py b/test/test_cuda/core/test_main_func.py
index 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, AutoTokenizer
from transformers.utils.versions import require_version
from 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_ut
from ...helpers import evaluate_accuracy, get_model_path
diff --git a/test/test_cuda/export/test_auto_round_format.py b/test/test_cuda/export/test_auto_round_format.py
index 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 transformers
from transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizer
from 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_infer
diff --git a/test/test_cuda/export/test_export.py b/test/test_cuda/export/test_export.py
index 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 version
from transformers import AutoConfig, AutoModelForCausalLM, AutoRoundConfig, AutoTokenizer
from 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_ut
from ...helpers import get_model_path, get_tiny_model, transformers_version
diff --git a/test/test_cuda/export/test_gguf.py b/test/test_cuda/export/test_gguf.py
index 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 version
from transformers import AutoModelForCausalLM, AutoTokenizer
from auto_round import AutoRound
-from auto_round.testing_utils import require_gguf
+from ...envs import require_gguf
from ...helpers import evaluate_accuracy, get_model_path, get_tiny_model, save_tiny_model, transformers_version
AUTO_ROUND_PATH = __file__.split("/")
diff --git a/test/test_cuda/integrations/test_transformers.py b/test/test_cuda/integrations/test_transformers.py
index 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_available
from ...helpers import get_model_path
if is_torch_available():
diff --git a/test/test_cuda/models/test_conv1d.py b/test/test_cuda/models/test_conv1d.py
index 8c1f654a3..deae349b3 100644
--- a/test/test_cuda/models/test_conv1d.py
+++ b/test/test_cuda/models/test_conv1d.py
@@ -6,8 +6,8 @@ import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from auto_round import AutoRound
-from auto_round.testing_utils import require_gptqmodel
+from ...envs import require_gptqmodel
from ...helpers import get_model_path, get_tiny_model, model_infer
diff --git a/test/test_cuda/models/test_diffusion.py b/test/test_cuda/models/test_diffusion.py
index 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 version
from PIL import Image
from 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_env
from ...helpers import transformers_version
diff --git a/test/test_cuda/models/test_support_vlms.py b/test/test_cuda/models/test_support_vlms.py
index 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 version
from PIL import Image
from 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_env
from ...helpers import transformers_version
AUTO_ROUND_PATH = __file__.split("/")
diff --git a/test/test_cuda/models/test_vlms.py b/test/test_cuda/models/test_vlms.py
index ad7fcaa11..3cbdc2a67 100644
--- a/test/test_cuda/models/test_vlms.py
+++ b/test/test_cuda/models/test_vlms.py
@@ -8,7 +8,7 @@ import requests
from PIL import Image
from 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_env
class TestAutoRound:
diff --git a/test/test_cuda/quantization/test_2_3bits.py b/test/test_cuda/quantization/test_2_3bits.py
index 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=E0401
from transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizer
from 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_051
from ...helpers import evaluate_accuracy, get_model_path, model_infer
diff --git a/test/test_cuda/quantization/test_mix_bits.py b/test/test_cuda/quantization/test_mix_bits.py
index 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 torch
from transformers import AutoModelForCausalLM, AutoRoundConfig, AutoTokenizer
from 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_path
diff --git a/test/test_cuda/quantization/test_mxfp_and_nvfp_quant.py b/test/test_cuda/quantization/test_mxfp_and_nvfp_quant.py
index 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_schemes
from auto_round.experimental import qmodules as ar_qmodules
from auto_round.export.export_to_autoround import qlinear_fp as ar_qlinear_fp
from auto_round.formats import AutoRoundExportFormat
-from auto_round.testing_utils import has_module
+from ...envs import has_module
from ...helpers import get_model_path
testing_schemes = [
diff --git a/test/test_cuda/quantization/test_mxfp_nvfp.py b/test/test_cuda/quantization/test_mxfp_nvfp.py
index 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 transformers
from transformers import AutoConfig, AutoModelForCausalLM, AutoTokenizer
from auto_round import AutoRound
-from auto_round.testing_utils import require_awq, require_optimum
+from ...envs import require_awq, require_optimum
from ...helpers import get_model_path, save_tiny_model
diff --git a/test/test_cuda/schemes/test_auto_scheme.py b/test/test_cuda/schemes/test_auto_scheme.py
index 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 AutoRoundConfig
from auto_round import AutoRound, AutoScheme
from auto_round.auto_scheme.utils import compute_avg_bits_for_model
-from auto_round.testing_utils import multi_card
from auto_round.utils import get_module
+from ...envs import multi_card
from ...helpers import evaluate_accuracy, get_model_path, get_tiny_model