[XPU] Add W8A8 FP8 linear kernel with multi-granularity quant support (#43645)

Signed-off-by: Chaojun,Zhang <chaojun.zhang@intel.com>
Signed-off-by: Chaojun Zhang <chaojun.zhang@intel.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This commit is contained in:
Chaojun Zhang
2026-07-04 18:10:01 +08:00
committed by GitHub
co-authored by Copilot
parent e7c9df9449
commit 1a308c449c
6 changed files with 188 additions and 12 deletions
@@ -13,6 +13,7 @@ from compressed_tensors.quantization import (
)
from compressed_tensors.transform import TransformConfig
from vllm.config import get_current_vllm_config_or_none
from vllm.distributed import (
get_tensor_model_parallel_rank,
get_tensor_model_parallel_world_size,
@@ -747,9 +748,18 @@ class CompressedTensorsConfig(QuantizationConfig):
act_quant_format = is_activation_quantization_format(format)
if act_quant_format:
if self._is_fp8_w8a8(weight_quant, input_quant):
is_fp8_w8a8_supported = self._check_scheme_supported(
CompressedTensorsW8A8Fp8.get_min_capability(), error=False
)
if current_platform.is_xpu():
# On XPU, --linear-backend xpu opts into W8A8 FP8
# linear kernel; otherwise default to W8A16.
config = get_current_vllm_config_or_none()
is_fp8_w8a8_supported = (
config is not None
and config.kernel_config.linear_backend == "xpu"
)
else:
is_fp8_w8a8_supported = self._check_scheme_supported(
CompressedTensorsW8A8Fp8.get_min_capability(), error=False
)
if is_fp8_w8a8_supported:
return CompressedTensorsW8A8Fp8(
weight_quant=weight_quant,