biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Misc] Fix quantization-related typos (#31116)

Browse Source 
Signed-off-by: c0de128 <kevin.mckay@outlook.com>
This commit is contained in:
Kevin McKay
2025-12-21 23:13:48 -06:00
committed by GitHub
parent 8c084de59d
commit ec58c10ce1
3 changed files with 7 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
tests/kernels/moe/modular_kernel_tools/common.py
View File

@@ -258,16 +258,16 @@ class Config:
f"{self.fe_supported_types()}."
)
# Check block quanization support
is_block_quatized = self.quant_block_shape is not None
if is_block_quatized and self.quant_dtype is None:
# Check block quantization support
is_block_quantized = self.quant_block_shape is not None
if is_block_quantized and self.quant_dtype is None:
return False, "No block quantization support."
if is_block_quatized and not self.is_block_quant_supported():
if is_block_quantized and not self.is_block_quant_supported():
return False, "Mismatched block quantization support."
# deep_gemm only works with block-quantized
if self.needs_deep_gemm() and not is_block_quatized:
if self.needs_deep_gemm() and not is_block_quantized:
return False, "Needs DeepGEMM but not block quantized."
# Check dependencies (turn into asserts?)