Reduce memory usage

Xia-Weiwen · May 11, 2024 · dd15734 · dd15734
1 parent 881b5fc
commit dd15734
Showing 1 changed file with 1 addition and 0 deletions.
diff --git a/bitsandbytes/backends/cpu_xpu_common.py b/bitsandbytes/backends/cpu_xpu_common.py
@@ -377,6 +377,7 @@ def quantize_4bit_impl(
             int(lowp_mode),
             -1,  # act_quant_mode. -1 means don't quant activation
         )
+        return torch.Tensor(), state
 
     return out, state