Add assertion when loading cpu and cuda kernel fails

2023-04-14 20:00:41 +08:00 · 2023-04-14 20:00:41 +08:00 · 630d0efd8b
parent bcc35f08b4
commit 630d0efd8b
1 changed files with 1 additions and 1 deletions
--- a/quantization.py
+++ b/quantization.py
@ -443,7 +443,7 @@ def quantize(model, weight_bit_width, use_quantization_cache=False, empty_init=F
    except:
        if kernels is None:  # CUDA kernels failed
            print("Cannot load cpu or cuda kernel, quantization failed:")
-            assert kernels is None
+            assert kernels is not None
        print("Cannot load cpu kernel, don't use quantized model on cpu.")

    current_device = model.device