Apply isort and black

Author: FKKimura

onecomp/quantized_model_loader.py

@@ -23,8 +23,8 @@
 from .quantizer.gptq.config import resolve_gptq_layer_group_size, resolve_gptq_layer_wbits
 from .quantizer.gptq.gptq_layer import GPTQLinear
 from .quantizer.onebit.onebit_layer import OneBitLinear
-from .utils.dtype import needs_bfloat16
 from .utils.device import get_default_device
+from .utils.dtype import needs_bfloat16
 from .utils.quant_config import get_quant_param
 
 logger = getLogger(__name__)

onecomp/quantizer/autobit/activation_stats.py

@@ -16,7 +16,7 @@
     get_blocks_and_inputs,
     move_kwargs_to_device,
 )
-from onecomp.utils.device import get_default_device, empty_cache
+from onecomp.utils.device import empty_cache, get_default_device
 
 
 def _find_head_modules(model, blocks):

onecomp/quantizer/gptq/_gptq.py

@@ -20,8 +20,8 @@
 from transformers import Conv1D
 
 from onecomp.quantizer._quantizer import QuantizationResult, Quantizer
-from onecomp.utils.quant_config import get_quant_param
 from onecomp.utils.device import empty_cache
+from onecomp.utils.quant_config import get_quant_param
 
 
 @dataclass

onecomp/runner.py

@@ -330,13 +330,8 @@ def check(self):
         device = self.model_config.device
         if is_mps_device(device):
             if self.multi_gpu:
-                raise ValueError(
-                    "multi_gpu is not supported on MPS device."
-                )
-            all_quantizers = (
-                self.quantizers if self.quantizers is not None
-                else [self.quantizer]
-            )
+                raise ValueError("multi_gpu is not supported on MPS device.")
+            all_quantizers = self.quantizers if self.quantizers is not None else [self.quantizer]
             for i, q in enumerate(all_quantizers):
                 label = f"quantizers[{i}]" if self.quantizers else "quantizer"
                 if isinstance(q, AutoBitQuantizer):
@@ -583,7 +578,9 @@ def auto_run(
                 enable_fused_groups=True,
             )
         qep_config = QEPConfig(device=device)
-        runner = cls(model_config=model_config, quantizer=quantizer, qep=qep, qep_config=qep_config)
+        runner = cls(
+            model_config=model_config, quantizer=quantizer, qep=qep, qep_config=qep_config
+        )
         runner.run()
 
         if evaluate:

onecomp/utils/perplexity.py

@@ -125,7 +125,9 @@ def calculate_perplexity(
     if stride is None:
         stride = max_length
     seq_len = encodings.input_ids.size(1)
-    use_cpu_accum = device.type == "mps" if isinstance(device, torch.device) else str(device).startswith("mps")
+    use_cpu_accum = (
+        device.type == "mps" if isinstance(device, torch.device) else str(device).startswith("mps")
+    )
     accum_device = torch.device("cpu") if use_cpu_accum else device
     nll_sum = torch.tensor(0.0, dtype=torch.float64, device=accum_device)
     n_tokens = 0