As of 21d4c48, the following aten ops are not implemented by XPU backend which are required by https://github.com/dvrogozh/optimum-quanto:

• __lshift__.Scalar, Add aten::lshift/rshift/prelu and thieir variants  intel/torch-xpu-ops#688
• aten::_weight_int8pack_mm (requires PYTORCH_ENABLE_XPU_FALLBACK=1)
• aten::_int_mm (requires PYTORCH_ENABLE_XPU_FALLBACK=1)
• aten::_convert_weight_to_int4pack (requires PYTORCH_ENABLE_XPU_FALLBACK=1), Add aten::_convert_weight_to_int4pack intel/torch-xpu-ops#1035

Support for XPU backend in quanto can be tried applying the following PRs:

• xpu: support sycl with torch.utils.cpp_extension APIs #132945
• feat: implement unpack sycl kernels with xpu extension huggingface/optimum-quanto#273

CC: @gujinghui @EikanWang @fengyuan14 @guangyey @jgong5

cc @gujinghui @EikanWang @fengyuan14 @guangyey