lib.cquantize_blockwise_fp32 mutates absmax

[cyr0930]

System Info

OS: Ubuntu 22.04.5 LTS
Python version: 3.11.11
GPU: A100-80GB
Driver version: 565.57.01
CUDA version: 12.7
bitsandbytes version: 0.45.5

Reproduction

a = torch.tensor([i / 10 for i in range(10)])
b = torch.tensor([i / 10 for i in range(10)])
functional.quantize_blockwise(a, absmax=b)

after that
b = tensor([0.9000, 0.1000, 0.2000, 0.3000, 0.4000, 0.5000, 0.6000, 0.7000, 0.8000, 0.9000])

Expected behavior

This should not mutate absmax (imo)

[cyr0930]

functional.quantize_blockwise calls lib.cquantize_blockwise_fp32.

I haven't tested lib.cquantize_blockwise_fp16 and lib.cquantize_blockwise_bf16.