[XPU] Fix FP8 block-scaled scheme selection on non-CUDA platforms (#43958)
Signed-off-by: Lai, Yejing <yejing.lai@intel.com> Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
This commit is contained in:
co-authored by
Kunshang Ji
parent
e1a5fc406b
commit
8dd8b6ed78
@@ -480,6 +480,7 @@ def test_compressed_tensors_fp8_block_enabled(vllm_runner):
|
|||||||
assert input_quant_op._forward_method in (
|
assert input_quant_op._forward_method in (
|
||||||
input_quant_op.forward_cuda,
|
input_quant_op.forward_cuda,
|
||||||
input_quant_op.forward_hip,
|
input_quant_op.forward_hip,
|
||||||
|
input_quant_op.forward_xpu,
|
||||||
)
|
)
|
||||||
|
|
||||||
llm.apply_model(check_model)
|
llm.apply_model(check_model)
|
||||||
|
|||||||
@@ -396,7 +396,7 @@ class CompressedTensorsConfig(QuantizationConfig):
|
|||||||
)
|
)
|
||||||
return supported
|
return supported
|
||||||
else:
|
else:
|
||||||
return False
|
return not match_exact
|
||||||
|
|
||||||
@staticmethod
|
@staticmethod
|
||||||
def _is_nvfp4_format(quant_args: QuantizationArgs):
|
def _is_nvfp4_format(quant_args: QuantizationArgs):
|
||||||
|
|||||||
Reference in New Issue
Block a user