fix bug

intel · Jun 4, 2024 · 04a5149 · 04a5149
1 parent 4d528f3
commit 04a5149
Showing 1 changed file with 2 additions and 2 deletions.
diff --git a/auto_round/autoround.py b/auto_round/autoround.py
@@ -738,7 +738,7 @@ def quant_layer(self, layer_name, inputs, q_inputs=None, device=torch.device("cp
         mse_loss = torch.nn.MSELoss().to(device)
         scaler = self.get_scaler()  # pylint: disable=assignment-from-none
         init_loss = None
-        best_v, best_min_scale, best_max_scale = torch.tensor(0), torch.tensor(0), torch.tensor(0)
+        best_v, best_min_scale, best_max_scale = torch.tensor(0), torch.tensor(1.0), torch.tensor(1.0)
         gradient_accumulate_steps = self.train_bs  ##Force to low gpu
         train_bs = 1  ##Force to low gpu
         pick_samples = train_bs * gradient_accumulate_steps
@@ -861,7 +861,7 @@ def quant_block(self, block, input_ids, input_others, q_input=None, device=torch
         mse_loss = torch.nn.MSELoss().to(device)
         scaler = self.get_scaler()  # pylint: disable=assignment-from-none
         init_loss = None
-        best_v, best_min_scale, best_max_scale = torch.tensor(0), torch.tensor(0), torch.tensor(0)
+        best_v, best_min_scale, best_max_scale = torch.tensor(0), torch.tensor(1.0), torch.tensor(1.0)
         for i in range(self.iters):
             total_loss = 0
             if self.sampler == "rand":