refactor(model): remove explicit device_type parameter from amp decorators

llava/model/qlinear_te.py

@@ -98,7 +98,7 @@ class QLinearTE(nn.Linear):
 
 class QuantLinearTE(Function):
     @staticmethod
-    @amp.custom_fwd(cast_inputs=torch.bfloat16, device_type='cuda')
+    @amp.custom_fwd(cast_inputs=torch.bfloat16)
     def forward(ctx, input, weight, bias, args, layer_name):
 
         time_bench = os.getenv("TIME_BENCH")
@@ -149,7 +149,7 @@ class QuantLinearTE(Function):
         return fc_output
 
     @staticmethod
-    @amp.custom_bwd(device_type='cuda')
+    @amp.custom_bwd
     def backward(ctx, grad_output):
         Qinput_t, Iscale, Qweight_t, Wscale, bias, args, layer_name = ctx.saved