[pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

Author: pre-commit-ci[bot]

neural_compressor/common/utils/__init__.py

@@ -13,9 +13,8 @@
 # limitations under the License.
 
 from neural_compressor.common.utils.constants import *
-from neural_compressor.common.utils.logger import 
+from neural_compressor.common.utils.logger import
 from neural_compressor.common.utils.save_load import save_qconfig, load_qconfig
 
 # ! Put the following `utility` import after the `logger` import as `utility` used `logger`
 from neural_compressor.common.utils.utility import *
-

neural_compressor/common/utils/save_load.py

@@ -15,8 +15,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import os
 import json
+import os
 
 
 def save_qconfig(config_mapping, qconfig_file_path):
@@ -35,8 +35,9 @@ def save_qconfig(config_mapping, qconfig_file_path):
     with open(qconfig_file_path, "w") as f:
         json.dump(per_op_qconfig, f, indent=4)
 
+
 def load_qconfig(qconfig_file_path, config_name_mapping):
-    """reload config mapping from json file.
+    """Reload config mapping from json file.
 
     Args:
         qconfig_file_path (str): path to saved json file.

neural_compressor/torch/algorithms/habana_fp8/modules.py

@@ -301,7 +301,6 @@ def __init__(self, org_module, dtype) -> None:
             if org_module.bias is not None:
                 self.bias.data.copy_(org_module.bias.data.type(self.out_dtype))
 
-
     def forward(self, inp):
         assert inp.shape[-1] == self.in_features, "GEMM not possible"
         org_middle_shape = inp.shape[1:-1]
@@ -638,6 +637,7 @@ def forward(self, inp):
             False,
         )
         from deepspeed import comm as dist
+
         if self.mp_group is not None:
             dist.inference_all_reduce(out, group=self.mp_group)
         if self.bias is not None:
@@ -762,6 +762,7 @@ def forward(self, inp):
             False,
         )
         from deepspeed import comm as dist
+
         if self.mp_group is not None:
             dist.inference_all_reduce(out, group=self.mp_group)
         if self.bias is not None:

neural_compressor/torch/algorithms/habana_fp8/save_load.py

@@ -18,8 +18,8 @@
 import habana_frameworks.torch.core as htcore
 import torch
 
-from neural_compressor.torch.utils import logger, WEIGHT_NAME, QCONFIG_NAME
-from neural_compressor.common.utils import save_qconfig, load_qconfig
+from neural_compressor.common.utils import load_qconfig, save_qconfig
+from neural_compressor.torch.utils import QCONFIG_NAME, WEIGHT_NAME, logger
 
 from .fp8_quant import FP8_DTYPE, dtype_mapping
 from .modules import (  # fp32; dynamic modules
@@ -62,7 +62,7 @@ def load(model, output_dir="./saved_results"):
     qmodel_file_path = os.path.join(os.path.abspath(os.path.expanduser(output_dir)), WEIGHT_NAME)
     stat_dict = torch.load(qmodel_file_path)
     import fp8_convert
-    
+
     for (op_name, op_type), op_qconfig in model.qconfig.items():
         dtype = op_qconfig.w_dtype
         choice = 1 if dtype == "fp8_e4m3" else 0
@@ -90,7 +90,7 @@ def load(model, output_dir="./saved_results"):
         set_module(model, op_name, module)
         htcore.mark_step()
     model.load_state_dict(stat_dict, assign=True)
-    model.to('hpu')
+    model.to("hpu")
     htcore.mark_step()
     logger.info("Quantized model loading successful.")
     return model

neural_compressor/torch/quantization/algorithm_entry.py

@@ -21,12 +21,12 @@
 from neural_compressor.common.utils import AWQ, FP8_QUANT, GPTQ, HQQ, RTN, STATIC_QUANT, TEQ
 from neural_compressor.torch.quantization import (
     AWQConfig,
+    FP8Config,
     GPTQConfig,
     HQQConfig,
     RTNConfig,
-    TEQConfig,
     StaticQuantConfig,
-    FP8Config,
+    TEQConfig,
 )
 from neural_compressor.torch.utils import logger, register_algo
 

neural_compressor/torch/quantization/load_entry.py

@@ -15,13 +15,15 @@
 import json
 import os
 
-from neural_compressor.common.utils import load_qconfig  # unified namespace
 from neural_compressor.common.utils import FP8_QUANT  # unified namespace
+from neural_compressor.common.utils import load_qconfig  # unified namespace
 from neural_compressor.torch.quantization.config import FP8Config
+
 config_name_mapping = {
     FP8_QUANT: FP8Config,
 }
 
+
 def load(model, output_dir="./saved_results"):
     qconfig_file_path = os.path.join(os.path.abspath(os.path.expanduser(output_dir)), "qconfig.json")
     config_mapping = load_qconfig(qconfig_file_path, config_name_mapping)