fix typo (code dupe cute/paste)

quantize.py

@@ -474,86 +474,6 @@ def forward(self, indices: torch.Tensor) -> torch.Tensor:
 ### HQQ
 
 class WeightOnlyInt4HqqQuantHandler:
-    def __init__(self, mod, group_size):
-        self.mod = mod
-        self.groupsize = group_size
-
-    def _create_quantized_state_dict(self):
-        from hqq.core.quantize import Quantizer  # TODO maybe torchao
-
-        for m in self.mod.modules():
-            for name, child in m.named_children():
-                if isinstance(child, torch.nn.Linear):
-                    child.weight = torch.nn.Parameter(
-                        Quantizer.dequantize(
-                            *Quantizer.quantize(
-                                child.weight,
-                                nbits=4,
-                                group_size=self.groupsize,
-                                axis=1,
-                            )
-                        )
-                    )
-
-        return WeightOnlyInt4QuantHandler(self.mod, self.groupsize).create_quantized_state_dict()
-
-    def _convert_for_runtime(self):
-        return WeightOnlyInt4GPTQQuantHandler(self.mod, self.groupsize).convert_for_runtime(use_cuda=True)
-
-
-    class WeightOnlyInt4HqqQuantHandler:
-    def __init__(self, mod, groupsize):
-        self.mod = mod
-        self.groupsize = groupsize
-
-    def _create_quantized_state_dict(self):
-        from hqq.core.quantize import Quantizer  # TODO maybe torchao
-
-        for m in self.mod.modules():
-            for name, child in m.named_children():
-                if isinstance(child, torch.nn.Linear):
-                    child.weight = torch.nn.Parameter(
-                        Quantizer.dequantize(
-                            *Quantizer.quantize(
-                                child.weight,
-                                nbits=4,
-                                group_size=self.groupsize,
-                                axis=1,
-                            )
-                        )
-                    )
-
-        return WeightOnlyInt4QuantHandler(self.mod, self.groupsize).create_quantized_state_dict()
-
-    def _convert_for_runtime(self):
-        return WeightOnlyInt4GPTQQuantHandler(self.mod, self.groupsize).convert_for_runtime(use_cuda=True)
-    class WeightOnlyInt4HqqQuantHandler:
-    def __init__(self, mod, groupsize):
-        self.mod = mod
-        self.groupsize = groupsize
-
-    def _create_quantized_state_dict(self):
-        from hqq.core.quantize import Quantizer  # TODO maybe torchao
-
-        for m in self.mod.modules():
-            for name, child in m.named_children():
-                if isinstance(child, torch.nn.Linear):
-                    child.weight = torch.nn.Parameter(
-                        Quantizer.dequantize(
-                            *Quantizer.quantize(
-                                child.weight,
-                                nbits=4,
-                                group_size=self.groupsize,
-                                axis=1,
-                            )
-                        )
-                    )
-
-        return WeightOnlyInt4QuantHandler(self.mod, self.groupsize).create_quantized_state_dict()
-
-    def _convert_for_runtime(self):
-        return WeightOnlyInt4GPTQQuantHandler(self.mod, self.groupsize).convert_for_runtime(use_cuda=True)
-    class WeightOnlyInt4HqqQuantHandler:
     def __init__(self, mod, groupsize):
         self.mod = mod
         self.groupsize = groupsize