Merge branch 'kaihui/save_and_load' of https://github.com/intel/neural-compressor into kaihui/save_and_load

Author: Kaihui-intel

test/3x/torch/quantization/test_smooth_quant.py

@@ -1,8 +1,8 @@
 import copy
+import shutil
 
 import pytest
 import torch
-import shutil
 
 from neural_compressor.torch.quantization import SmoothQuantConfig, get_default_sq_config, quantize
 from neural_compressor.torch.utils import is_ipex_available

test/3x/torch/quantization/test_static_quant.py

@@ -1,8 +1,8 @@
 import copy
+import shutil
 
 import pytest
 import torch
-import shutil
 
 from neural_compressor.torch.quantization import (
     StaticQuantConfig,

test/3x/torch/quantization/weight_only/test_autoround.py

@@ -1,9 +1,9 @@
 import copy
+import shutil
 
 import pytest
 import torch
 import transformers
-import shutil
 
 from neural_compressor.torch.algorithms.weight_only.autoround import AutoRoundQuantizer, get_autoround_default_run_fn
 from neural_compressor.torch.quantization import (
@@ -38,7 +38,7 @@ def setup_class(self):
 
     def teardown_class(self):
         shutil.rmtree("saved_results", ignore_errors=True)
-    
+
     def setup_method(self, method):
         logger.info(f"Running TestAutoRound test: {method.__name__}")
 

test/3x/torch/quantization/weight_only/test_awq.py

@@ -1,9 +1,9 @@
 import copy
+import shutil
 
 import pytest
 import torch
 import transformers
-import shutil
 
 from neural_compressor.common import Logger
 

test/3x/torch/quantization/weight_only/test_gptq.py

@@ -1,9 +1,9 @@
 import copy
+import shutil
 
 import pytest
 import torch
 import transformers
-import shutil
 
 from neural_compressor.torch.algorithms.weight_only.modules import WeightOnlyLinear
 from neural_compressor.torch.quantization import (
@@ -41,7 +41,6 @@ def setup_class(self):
     def teardown_class(self):
         shutil.rmtree("saved_results", ignore_errors=True)
 
-
     def test_accuracy_improvement(self):
         # test_default_rtn_config
         model = copy.deepcopy(self.tiny_gptj)

test/3x/torch/quantization/weight_only/test_rtn.py

@@ -1,9 +1,9 @@
 import copy
+import shutil
 
 import pytest
 import torch
 import transformers
-import shutil
 
 from neural_compressor.torch.algorithms.weight_only.modules import WeightOnlyLinear
 from neural_compressor.torch.quantization import (
@@ -35,7 +35,6 @@ def setup_class(self):
     def teardown_class(self):
         shutil.rmtree("saved_results", ignore_errors=True)
 
-
     # TODO: (4, True, 32, 0), group_dim=0, format not supported
     @pytest.mark.parametrize(
         "bits, use_sym, group_size, group_dim",

test/3x/torch/quantization/weight_only/test_teq.py

@@ -1,9 +1,9 @@
 import copy
+import shutil
 import unittest
 
 import torch
 import transformers
-import shutil
 
 from neural_compressor.common import logger
 from neural_compressor.torch.quantization import convert, prepare, quantize
@@ -113,7 +113,6 @@ def setUpClass(self):
     def tearDownClass(self):
         shutil.rmtree("saved_results", ignore_errors=True)
 
-    
     def test_teq(self):
         test_input = torch.ones([1, 512], dtype=torch.long)
         model = copy.deepcopy(self.gptj)