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 11, 03:45 UTC
[Bugfix][ROCm] Fix typo: is_linear_fp8_enaled -> is_linear_fp8_enabled (#31109)
Signed-off-by: c0de128 <kevin.mckay@outlook.com> Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Kevin McKay committed
Dec 21, 2025 at 23:14 UTC
cf8eed7bef52293a9a6706a2953374b24dff2c81
5 files changed
+5
-5
vllm/_aiter_ops.py
+1
-1
@@ -761,7 +761,7 @@ class rocm_aiter_ops:
761
762
@classmethod
763
@if_aiter_supported
764
- def is_linear_fp8_enaled(cls) -> bool:
764
+ def is_linear_fp8_enabled(cls) -> bool:
765
return cls.is_linear_enabled()
766
767
@classmethod
vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_fp8.py
+1
-1
@@ -61,7 +61,7 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
61
)
62
63
self.cutlass_block_fp8_supported = cutlass_block_fp8_supported()
64
- self.use_aiter_and_is_supported = rocm_aiter_ops.is_linear_fp8_enaled()
64
+ self.use_aiter_and_is_supported = rocm_aiter_ops.is_linear_fp8_enabled()
65
66
if self.weight_block_size is not None:
67
assert not self.is_static_input_scheme
vllm/model_executor/layers/quantization/fp8.py
+1
-1
@@ -414,7 +414,7 @@ class Fp8LinearMethod(LinearMethodBase):
414
if vllm_is_batch_invariant():
415
self.use_marlin = False
416
417
- self.use_aiter_and_is_supported = rocm_aiter_ops.is_linear_fp8_enaled()
417
+ self.use_aiter_and_is_supported = rocm_aiter_ops.is_linear_fp8_enabled()
418
self.use_deep_gemm = is_deep_gemm_supported()
419
420
self.weight_block_size = self.quant_config.weight_block_size
vllm/model_executor/layers/quantization/input_quant_fp8.py
+1
-1
@@ -51,7 +51,7 @@ class QuantFP8(CustomOp):
51
self.column_major_scales = column_major_scales
52
self.use_ue8m0 = use_ue8m0
53
54
- self.use_aiter = rocm_aiter_ops.is_linear_fp8_enaled()
54
+ self.use_aiter = rocm_aiter_ops.is_linear_fp8_enabled()
55
56
self.is_group_quant = group_shape.is_per_group()
57
if self.is_group_quant:
vllm/platforms/rocm.py
+1
-1
@@ -408,7 +408,7 @@ class RocmPlatform(Platform):
408
parallel_config = vllm_config.parallel_config
409
is_eager_execution = compilation_config == CUDAGraphMode.NONE
410
use_aiter_rms_norm = rocm_aiter_ops.is_rmsnorm_enabled()
411
- use_aiter_fp8_linear = rocm_aiter_ops.is_linear_fp8_enaled()
411
+ use_aiter_fp8_linear = rocm_aiter_ops.is_linear_fp8_enabled()
412
413
if compilation_config.cudagraph_mode.has_full_cudagraphs():
414
# decode context parallel does not support full cudagraphs