A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 03:30 UTC
[Bugfix][CI/CD][CPU] Fix CPU CI tests (#20383)
Signed-off-by: jiang1.li <jiang1.li@intel.com>
Li, Jiang committed
Jul 3, 2025 at 11:11 UTC
0ec3779df74ff68ab920856234e2a1aafc21f1a1
1 file changed
+7
-6
vllm/model_executor/layers/quantization/utils/fp8_utils.py
+7
-6
@@ -201,12 +201,13 @@ def apply_w8a8_block_fp8_linear_fake(
201
return torch.empty(output_shape, dtype=input.dtype, device=input.device)
202
203
204
-direct_register_custom_op(
205
- op_name="apply_w8a8_block_fp8_linear",
206
- op_func=apply_w8a8_block_fp8_linear,
207
- mutates_args=[],
208
- fake_impl=apply_w8a8_block_fp8_linear_fake,
209
-)
204
+if not current_platform.is_cpu():
205
+ direct_register_custom_op(
206
+ op_name="apply_w8a8_block_fp8_linear",
207
+ op_func=apply_w8a8_block_fp8_linear,
208
+ mutates_args=[],
209
+ fake_impl=apply_w8a8_block_fp8_linear_fake,
210
+ )
211
212
213
def input_to_float8(