Make code cleaner

awq/quantize/quantizer.py

@@ -206,7 +206,7 @@ class AwqQuantizer:
         """
         Compute loss and select best scales
 
-        L(s) = ||Q(W \cdot s) (s^{-1} \cdot X) - W \cdot X||
+        L(s) = || Q(W * s) (s^-1 * X) - W * X ||
         Q: weight quantization function | pseudo_quantize_tensor(W * s)
         X: inputs from calib dataset    | X
         W: original weights in FP16     | layer
@@ -219,16 +219,21 @@ class AwqQuantizer:
         best_error = float('inf')
 
         org_sd = {k: v.cpu() for k, v in previous_layer.state_dict().items()}
+        
+        device = x.device
+        x_max = x_max.view(-1).to(device)
+        w_max = w_max.view(-1).to(device)
+        
         for ratio in range(n_grid):
             # create new scales
-            ratio = ratio * 1 / n_grid
-            scales = (x_max.pow(ratio) / w_max.pow(1-ratio)).clamp(min=1e-4).view(-1)
+            ratio = ratio / n_grid
+            scales = (x_max.pow(ratio) / w_max.pow(1-ratio)).clamp(min=1e-4)
             scales = scales / (scales.max() * scales.min()).sqrt()
             
-            # multiply scale and quantize
+            scales_view = scales.view(1, -1).to(device)
             for fc in linears2scale:
-                fc.weight.mul_(scales.view(1, -1).to(fc.weight.device))
-                fc.weight.data = self.pseudo_quantize_tensor(fc.weight.data) / (scales.view(1, -1))
+                fc.weight.mul_(scales_view)
+                fc.weight.data = self.pseudo_quantize_tensor(fc.weight.data) / scales_view
 
             out = previous_layer(x, **kwargs)
             if isinstance(out, tuple):
@@ -237,22 +242,21 @@ class AwqQuantizer:
             # measure loss and check if better than best
             loss = (org_out - out).float().pow(2).mean().item() # NOTE: float prevents overflow
             history.append(loss)
-            is_best = loss < best_error
-            if is_best:
+            if loss < best_error:
                 best_error = loss
                 best_ratio = ratio
-                best_scales = scales
+                best_scales = scales.clone()
             previous_layer.load_state_dict(org_sd)
 
         if best_ratio == -1:
             logging.debug(history)
             raise Exception
 
-        best_scales = best_scales.view(-1)
         assert torch.isnan(best_scales).sum() == 0, best_scales
 
         return best_scales.detach()
 
+
     def init_quant(self, n_samples=128, seqlen=512):
         layers = self.get_model_layers(self.model)